Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
/**
2024-10-17 18:59:52 +00:00
* llama . cpp - commit 3f 1 ae2e32cde00c39b96be6d01c2997c29bae555 - do not edit this file
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
*
* MIT License
*
* Copyright ( c ) 2023 - 2024 The ggml authors
*
* Permission is hereby granted , free of charge , to any person obtaining a copy
* of this software and associated documentation files ( the " Software " ) , to deal
* in the Software without restriction , including without limitation the rights
* to use , copy , modify , merge , publish , distribute , sublicense , and / or sell
* copies of the Software , and to permit persons to whom the Software is
* furnished to do so , subject to the following conditions :
*
* The above copyright notice and this permission notice shall be included in all
* copies or substantial portions of the Software .
*
* THE SOFTWARE IS PROVIDED " AS IS " , WITHOUT WARRANTY OF ANY KIND , EXPRESS OR
* IMPLIED , INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY ,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT . IN NO EVENT SHALL THE
* AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM , DAMAGES OR OTHER
* LIABILITY , WHETHER IN AN ACTION OF CONTRACT , TORT OR OTHERWISE , ARISING FROM ,
* OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
* SOFTWARE .
*/
# include "llama-sampling.h"
2024-10-17 18:59:52 +00:00
# include "llama-vocab.h"
# include "llama-grammar.h"
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
# include <algorithm>
2024-10-17 18:59:52 +00:00
# include <cassert>
# include <cfloat>
# include <chrono>
# include <cmath>
# include <cstdlib>
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
# include <cstring>
# include <ctime>
# include <numeric>
2024-10-17 18:59:52 +00:00
# include <random>
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
# include <unordered_map>
2024-10-17 18:59:52 +00:00
static int llama_sample_dist ( llama_token_data_array * cur_p , std : : mt19937 & rng ) {
// iterator for the probabilities
# ifdef __GNUC__
# pragma GCC diagnostic push
# pragma GCC diagnostic ignored "-Wunused-local-typedefs"
# endif
struct probs_iterator {
typedef std : : input_iterator_tag iterator_category ;
typedef float value_type ;
typedef float * pointer ;
typedef float & reference ;
typedef ptrdiff_t difference_type ;
const llama_token_data * data ;
bool operator = = ( const probs_iterator & other ) const { return data = = other . data ; }
bool operator ! = ( const probs_iterator & other ) const { return data ! = other . data ; }
const float & operator * ( ) const { return data - > p ; }
probs_iterator & operator + + ( ) { + + data ; return * this ; }
probs_iterator operator + + ( int ) { probs_iterator tmp = * this ; + + data ; return tmp ; }
} ;
# ifdef __GNUC__
# pragma GCC diagnostic pop
# endif
std : : discrete_distribution < int > dist ( probs_iterator { cur_p - > data } , probs_iterator { cur_p - > data + cur_p - > size } ) ;
return dist ( rng ) ;
}
/*
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
static void llama_log_softmax ( float * array , size_t size ) {
float max_l = * std : : max_element ( array , array + size ) ;
float sum = 0.f ;
for ( size_t i = 0 ; i < size ; + + i ) {
float p = expf ( array [ i ] - max_l ) ;
sum + = p ;
array [ i ] = p ;
}
for ( size_t i = 0 ; i < size ; + + i ) {
array [ i ] = logf ( array [ i ] / sum ) ;
}
}
2024-10-17 18:59:52 +00:00
*/
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
2024-10-17 18:59:52 +00:00
static void llama_sampler_softmax_impl ( llama_token_data_array * cur_p ) {
GGML_ASSERT ( cur_p - > size > 0 ) ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
// Sort the logits in descending order
2024-10-17 18:59:52 +00:00
if ( ! cur_p - > sorted ) {
std : : sort ( cur_p - > data , cur_p - > data + cur_p - > size , [ ] ( const llama_token_data & a , const llama_token_data & b ) {
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
return a . logit > b . logit ;
} ) ;
2024-10-17 18:59:52 +00:00
cur_p - > sorted = true ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
}
2024-10-17 18:59:52 +00:00
float max_l = cur_p - > data [ 0 ] . logit ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
float cum_sum = 0.0f ;
2024-10-17 18:59:52 +00:00
for ( size_t i = 0 ; i < cur_p - > size ; + + i ) {
float p = expf ( cur_p - > data [ i ] . logit - max_l ) ;
cur_p - > data [ i ] . p = p ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
cum_sum + = p ;
}
2024-10-17 18:59:52 +00:00
for ( size_t i = 0 ; i < cur_p - > size ; + + i ) {
cur_p - > data [ i ] . p / = cum_sum ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
}
}
2024-10-17 18:59:52 +00:00
static void llama_sampler_top_k_impl ( llama_token_data_array * cur_p , int32_t k ) {
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
// TODO: move bucket sort to separate function so that top_p/tail_free/typical/softmax first is equally fast
2024-10-17 18:59:52 +00:00
// if (k >= (int32_t)cur_p->size) {
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
// return;
// }
if ( k < = 0 ) {
2024-10-17 18:59:52 +00:00
k = cur_p - > size ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
}
2024-10-17 18:59:52 +00:00
k = std : : min ( k , ( int ) cur_p - > size ) ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
// Sort scores in descending order
2024-10-17 18:59:52 +00:00
if ( ! cur_p - > sorted ) {
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
auto comp = [ ] ( const llama_token_data & a , const llama_token_data & b ) {
return a . logit > b . logit ;
} ;
if ( k < = 128 ) {
2024-10-17 18:59:52 +00:00
std : : partial_sort ( cur_p - > data , cur_p - > data + k , cur_p - > data + cur_p - > size , comp ) ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
} else {
constexpr int nbuckets = 128 ;
constexpr float bucket_low = - 10.0f ;
constexpr float bucket_high = 10.0f ;
constexpr float bucket_scale = nbuckets / ( bucket_high - bucket_low ) ;
constexpr float bucket_inter = - bucket_low * bucket_scale ;
2024-10-17 18:59:52 +00:00
std : : vector < int > bucket_idx ( cur_p - > size ) ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
std : : vector < int > histo ( nbuckets , 0 ) ;
2024-10-17 18:59:52 +00:00
for ( int i = 0 ; i < ( int ) cur_p - > size ; + + i ) {
const float val = cur_p - > data [ i ] . logit ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
int ib = int ( bucket_scale * val + bucket_inter ) ; //nbuckets * (val - bucket_low) / (bucket_high - bucket_low);
ib = std : : max ( 0 , std : : min ( nbuckets - 1 , ib ) ) ;
bucket_idx [ i ] = ib ;
+ + histo [ ib ] ;
}
int nhave = 0 ;
int ib = nbuckets - 1 ;
for ( ; ib > = 0 ; - - ib ) {
nhave + = histo [ ib ] ;
2024-10-17 18:59:52 +00:00
if ( nhave > = k ) {
break ;
}
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
}
std : : vector < llama_token_data > tmp_tokens ( nhave ) ;
2024-10-17 18:59:52 +00:00
auto * ptr = tmp_tokens . data ( ) ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
std : : vector < llama_token_data * > bucket_ptrs ;
bucket_ptrs . reserve ( nbuckets - ib ) ;
for ( int j = nbuckets - 1 ; j > = ib ; - - j ) {
bucket_ptrs . push_back ( ptr ) ;
ptr + = histo [ j ] ;
}
2024-10-17 18:59:52 +00:00
for ( int i = 0 ; i < ( int ) cur_p - > size ; + + i ) {
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
int j = bucket_idx [ i ] ;
if ( j > = ib ) {
2024-10-17 18:59:52 +00:00
* bucket_ptrs [ nbuckets - 1 - j ] + + = cur_p - > data [ i ] ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
}
}
ptr = tmp_tokens . data ( ) ;
int ndone = 0 ;
for ( int j = nbuckets - 1 ; j > ib ; - - j ) {
std : : sort ( ptr , ptr + histo [ j ] , comp ) ;
ptr + = histo [ j ] ;
ndone + = histo [ j ] ;
}
std : : partial_sort ( ptr , ptr + k - ndone , ptr + histo [ ib ] , comp ) ;
2024-10-17 18:59:52 +00:00
std : : memcpy ( cur_p - > data , tmp_tokens . data ( ) , k * sizeof ( llama_token_data ) ) ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
}
2024-10-17 18:59:52 +00:00
cur_p - > sorted = true ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
}
2024-10-17 18:59:52 +00:00
cur_p - > size = k ;
}
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
2024-10-17 18:59:52 +00:00
static uint32_t get_rng_seed ( uint32_t seed ) {
if ( seed = = LLAMA_DEFAULT_SEED ) {
// use system clock if std::random_device is not a true RNG
static bool is_rd_prng = std : : random_device ( ) . entropy ( ) = = 0 ;
if ( is_rd_prng ) {
return ( uint32_t ) std : : chrono : : system_clock : : now ( ) . time_since_epoch ( ) . count ( ) ;
}
std : : random_device rd ;
return rd ( ) ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
}
2024-10-17 18:59:52 +00:00
return seed ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
}
2024-10-17 18:59:52 +00:00
// llama_sampler API
const char * llama_sampler_name ( const struct llama_sampler * smpl ) {
if ( ! smpl - > iface ) {
return " (null) " ;
}
return smpl - > iface - > name ( smpl ) ;
}
void llama_sampler_accept ( struct llama_sampler * smpl , llama_token token ) {
if ( smpl - > iface - > accept ) {
smpl - > iface - > accept ( smpl , token ) ;
}
}
void llama_sampler_apply ( struct llama_sampler * smpl , struct llama_token_data_array * cur_p ) {
GGML_ASSERT ( smpl - > iface - > apply ) ;
smpl - > iface - > apply ( smpl , cur_p ) ;
}
void llama_sampler_reset ( struct llama_sampler * smpl ) {
if ( smpl - > iface - > reset ) {
smpl - > iface - > reset ( smpl ) ;
}
}
struct llama_sampler * llama_sampler_clone ( const struct llama_sampler * smpl ) {
if ( smpl - > iface - > clone ) {
return smpl - > iface - > clone ( smpl ) ;
}
if ( smpl - > ctx = = nullptr ) {
return new llama_sampler {
/* .iface = */ smpl - > iface ,
/* .ctx = */ nullptr ,
} ;
}
GGML_ABORT ( " the sampler does not support cloning " ) ;
}
void llama_sampler_free ( struct llama_sampler * smpl ) {
if ( smpl = = nullptr ) {
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
return ;
}
2024-10-17 18:59:52 +00:00
if ( smpl - > iface - > free ) {
smpl - > iface - > free ( smpl ) ;
}
delete smpl ;
}
llama_token llama_sampler_sample ( struct llama_sampler * smpl , struct llama_context * ctx , int32_t idx ) {
const auto * logits = llama_get_logits_ith ( ctx , idx ) ;
const int n_vocab = llama_n_vocab ( llama_get_model ( ctx ) ) ;
// TODO: do not allocate each time
std : : vector < llama_token_data > cur ;
cur . reserve ( n_vocab ) ;
for ( llama_token token_id = 0 ; token_id < n_vocab ; token_id + + ) {
cur . emplace_back ( llama_token_data { token_id , logits [ token_id ] , 0.0f } ) ;
}
llama_token_data_array cur_p = {
/* .data = */ cur . data ( ) ,
/* .size = */ cur . size ( ) ,
/* .selected = */ - 1 ,
/* .sorted = */ false ,
} ;
llama_sampler_apply ( smpl , & cur_p ) ;
GGML_ASSERT ( cur_p . selected > = 0 & & cur_p . selected < ( int32_t ) cur_p . size ) ;
auto token = cur_p . data [ cur_p . selected ] . id ;
llama_sampler_accept ( smpl , token ) ;
return token ;
}
// sampler chain
static const char * llama_sampler_chain_name ( const struct llama_sampler * /*smpl*/ ) {
return " chain " ;
}
static void llama_sampler_chain_accept ( struct llama_sampler * smpl , llama_token token ) {
auto * chain = ( llama_sampler_chain * ) smpl - > ctx ;
time_meas tm ( chain - > t_sample_us , chain - > params . no_perf ) ;
for ( auto * smpl : chain - > samplers ) {
llama_sampler_accept ( smpl , token ) ;
}
chain - > n_sample + + ;
}
static void llama_sampler_chain_apply ( struct llama_sampler * smpl , llama_token_data_array * cur_p ) {
auto * chain = ( llama_sampler_chain * ) smpl - > ctx ;
time_meas tm ( chain - > t_sample_us , chain - > params . no_perf ) ;
for ( auto * smpl : chain - > samplers ) {
llama_sampler_apply ( smpl , cur_p ) ;
}
}
static void llama_sampler_chain_reset ( struct llama_sampler * smpl ) {
auto * chain = ( llama_sampler_chain * ) smpl - > ctx ;
for ( auto * smpl : chain - > samplers ) {
llama_sampler_reset ( smpl ) ;
}
chain - > t_sample_us = 0 ;
chain - > n_sample = 0 ;
}
static struct llama_sampler * llama_sampler_chain_clone ( const struct llama_sampler * smpl ) {
const auto * chain_src = ( const llama_sampler_chain * ) smpl - > ctx ;
auto * result = llama_sampler_chain_init ( chain_src - > params ) ;
for ( auto * smpl : chain_src - > samplers ) {
llama_sampler_chain_add ( result , llama_sampler_clone ( smpl ) ) ;
}
return result ;
}
static void llama_sampler_chain_free ( struct llama_sampler * smpl ) {
auto * chain = ( llama_sampler_chain * ) smpl - > ctx ;
for ( auto * smpl : chain - > samplers ) {
llama_sampler_free ( smpl ) ;
}
delete chain ;
}
static struct llama_sampler_i llama_sampler_chain_i = {
/* .name = */ llama_sampler_chain_name ,
/* .accept = */ llama_sampler_chain_accept ,
/* .apply = */ llama_sampler_chain_apply ,
/* .reset = */ llama_sampler_chain_reset ,
/* .clone = */ llama_sampler_chain_clone ,
/* .free = */ llama_sampler_chain_free ,
} ;
struct llama_sampler * llama_sampler_chain_init ( struct llama_sampler_chain_params params ) {
return new llama_sampler {
/* .iface = */ & llama_sampler_chain_i ,
/* .ctx = */ new llama_sampler_chain {
/* .params = */ params ,
/* .samplers = */ { } ,
/* .t_sample_us = */ 0 ,
/* .n_sample = */ 0 ,
} ,
} ;
}
void llama_sampler_chain_add ( struct llama_sampler * chain , struct llama_sampler * smpl ) {
auto * p = ( llama_sampler_chain * ) chain - > ctx ;
p - > samplers . push_back ( smpl ) ;
}
struct llama_sampler * llama_sampler_chain_get ( const struct llama_sampler * chain , int32_t i ) {
const auto * p = ( const llama_sampler_chain * ) chain - > ctx ;
if ( i < 0 | | ( size_t ) i > = p - > samplers . size ( ) ) {
return nullptr ;
}
return p - > samplers [ i ] ;
}
struct llama_sampler * llama_sampler_chain_remove ( struct llama_sampler * chain , int32_t i ) {
auto * p = ( llama_sampler_chain * ) chain - > ctx ;
if ( i < 0 | | ( size_t ) i > = p - > samplers . size ( ) ) {
return nullptr ;
}
auto * result = p - > samplers [ i ] ;
p - > samplers . erase ( p - > samplers . begin ( ) + i ) ;
return result ;
}
int llama_sampler_chain_n ( const struct llama_sampler * chain ) {
const auto * p = ( const llama_sampler_chain * ) chain - > ctx ;
return p - > samplers . size ( ) ;
}
//
// samplers
//
// greedy
static const char * llama_sampler_greedy_name ( const struct llama_sampler * /*smpl*/ ) {
return " greedy " ;
}
static void llama_sampler_greedy_apply ( struct llama_sampler * /*smpl*/ , llama_token_data_array * cur_p ) {
cur_p - > selected = 0 ;
for ( size_t i = 1 ; i < cur_p - > size ; + + i ) {
if ( cur_p - > data [ i ] . logit > cur_p - > data [ cur_p - > selected ] . logit ) {
cur_p - > selected = i ;
}
}
}
static struct llama_sampler_i llama_sampler_greedy_i = {
/* .name = */ llama_sampler_greedy_name ,
/* .accept = */ nullptr ,
/* .apply = */ llama_sampler_greedy_apply ,
/* .reset = */ nullptr ,
/* .clone = */ nullptr ,
/* .free = */ nullptr ,
} ;
struct llama_sampler * llama_sampler_init_greedy ( ) {
return new llama_sampler {
/* .iface = */ & llama_sampler_greedy_i ,
/* .ctx = */ nullptr ,
} ;
}
// dist
struct llama_sampler_dist {
const uint32_t seed ;
uint32_t seed_cur ;
std : : mt19937 rng ;
} ;
static const char * llama_sampler_dist_name ( const struct llama_sampler * /*smpl*/ ) {
return " dist " ;
}
static void llama_sampler_dist_apply ( struct llama_sampler * smpl , llama_token_data_array * cur_p ) {
auto * ctx = ( llama_sampler_dist * ) smpl - > ctx ;
cur_p - > selected = llama_sample_dist ( cur_p , ctx - > rng ) ;
}
static struct llama_sampler * llama_sampler_dist_clone ( const struct llama_sampler * smpl ) {
const auto * ctx = ( const llama_sampler_dist * ) smpl - > ctx ;
auto * result = llama_sampler_init_dist ( ctx - > seed ) ;
// copy the state
{
auto * result_ctx = ( llama_sampler_dist * ) result - > ctx ;
result_ctx - > rng = ctx - > rng ;
}
return result ;
}
static void llama_sampler_dist_reset ( struct llama_sampler * smpl ) {
auto * ctx = ( llama_sampler_dist * ) smpl - > ctx ;
ctx - > seed_cur = get_rng_seed ( ctx - > seed ) ;
ctx - > rng . seed ( ctx - > seed_cur ) ;
}
static void llama_sampler_dist_free ( struct llama_sampler * smpl ) {
delete ( llama_sampler_dist * ) smpl - > ctx ;
}
static struct llama_sampler_i llama_sampler_dist_i = {
/* .name = */ llama_sampler_dist_name ,
/* .accept = */ nullptr ,
/* .apply = */ llama_sampler_dist_apply ,
/* .reset = */ llama_sampler_dist_reset ,
/* .clone = */ llama_sampler_dist_clone ,
/* .free = */ llama_sampler_dist_free ,
} ;
struct llama_sampler * llama_sampler_init_dist ( uint32_t seed ) {
auto seed_cur = get_rng_seed ( seed ) ;
return new llama_sampler {
/* .iface = */ & llama_sampler_dist_i ,
/* .ctx = */ new llama_sampler_dist {
/* .seed = */ seed ,
/* .seed_cur = */ seed_cur ,
/* .rng = */ std : : mt19937 ( seed_cur ) ,
} ,
} ;
}
// softmax
static const char * llama_sampler_softmax_name ( const struct llama_sampler * /*smpl*/ ) {
return " softmax " ;
}
static void llama_sampler_softmax_apply ( struct llama_sampler * /*smpl*/ , llama_token_data_array * cur_p ) {
llama_sampler_softmax_impl ( cur_p ) ;
}
static struct llama_sampler_i llama_sampler_softmax_i = {
/* .name = */ llama_sampler_softmax_name ,
/* .accept = */ nullptr ,
/* .apply = */ llama_sampler_softmax_apply ,
/* .reset = */ nullptr ,
/* .clone = */ nullptr ,
/* .free = */ nullptr ,
} ;
struct llama_sampler * llama_sampler_init_softmax ( ) {
return new llama_sampler {
/* .iface = */ & llama_sampler_softmax_i ,
/* .ctx = */ nullptr ,
} ;
}
// top-k
struct llama_sampler_top_k {
const int32_t k ;
} ;
static const char * llama_sampler_top_k_name ( const struct llama_sampler * /*smpl*/ ) {
return " top-k " ;
}
static void llama_sampler_top_k_apply ( struct llama_sampler * smpl , llama_token_data_array * cur_p ) {
const auto * ctx = ( llama_sampler_top_k * ) smpl - > ctx ;
llama_sampler_top_k_impl ( cur_p , ctx - > k ) ;
}
static struct llama_sampler * llama_sampler_top_k_clone ( const struct llama_sampler * smpl ) {
const auto * ctx = ( const llama_sampler_top_k * ) smpl - > ctx ;
return llama_sampler_init_top_k ( ctx - > k ) ;
}
static void llama_sampler_top_k_free ( struct llama_sampler * smpl ) {
delete ( llama_sampler_top_k * ) smpl - > ctx ;
}
static struct llama_sampler_i llama_sampler_top_k_i = {
/* .name = */ llama_sampler_top_k_name ,
/* .accept = */ nullptr ,
/* .apply = */ llama_sampler_top_k_apply ,
/* .reset = */ nullptr ,
/* .clone = */ llama_sampler_top_k_clone ,
/* .free = */ llama_sampler_top_k_free ,
} ;
struct llama_sampler * llama_sampler_init_top_k ( int32_t k ) {
return new llama_sampler {
/* .iface = */ & llama_sampler_top_k_i ,
/* .ctx = */ new llama_sampler_top_k {
/* .k = */ k ,
} ,
} ;
}
// top-p
struct llama_sampler_top_p {
const float p ;
const size_t min_keep ;
} ;
static const char * llama_sampler_top_p_name ( const struct llama_sampler * /*smpl*/ ) {
return " top-p " ;
}
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
2024-10-17 18:59:52 +00:00
static void llama_sampler_top_p_apply ( struct llama_sampler * smpl , llama_token_data_array * cur_p ) {
const auto * ctx = ( llama_sampler_top_p * ) smpl - > ctx ;
if ( ctx - > p > = 1.0f ) {
return ;
}
llama_sampler_softmax_impl ( cur_p ) ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
// Compute the cumulative probabilities
float cum_sum = 0.0f ;
2024-10-17 18:59:52 +00:00
size_t last_idx = cur_p - > size ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
2024-10-17 18:59:52 +00:00
for ( size_t i = 0 ; i < cur_p - > size ; + + i ) {
cum_sum + = cur_p - > data [ i ] . p ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
// Check if the running sum is at least p or if we have kept at least min_keep tokens
// we set the last index to i+1 to indicate that the current iterate should be included in the set
2024-10-17 18:59:52 +00:00
if ( cum_sum > = ctx - > p & & i + 1 > = ctx - > min_keep ) {
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
last_idx = i + 1 ;
break ;
}
}
// Resize the output vector to keep only the top-p tokens
2024-10-17 18:59:52 +00:00
cur_p - > size = last_idx ;
}
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
2024-10-17 18:59:52 +00:00
static struct llama_sampler * llama_sampler_top_p_clone ( const struct llama_sampler * smpl ) {
const auto * ctx = ( const llama_sampler_top_p * ) smpl - > ctx ;
return llama_sampler_init_top_p ( ctx - > p , ctx - > min_keep ) ;
}
static void llama_sampler_top_p_free ( struct llama_sampler * smpl ) {
delete ( llama_sampler_top_p * ) smpl - > ctx ;
}
static struct llama_sampler_i llama_sampler_top_p_i = {
/* .name = */ llama_sampler_top_p_name ,
/* .accept = */ nullptr ,
/* .apply = */ llama_sampler_top_p_apply ,
/* .reset = */ nullptr ,
/* .clone = */ llama_sampler_top_p_clone ,
/* .free = */ llama_sampler_top_p_free ,
} ;
struct llama_sampler * llama_sampler_init_top_p ( float p , size_t min_keep ) {
return new llama_sampler {
/* .iface = */ & llama_sampler_top_p_i ,
/* .ctx = */ new llama_sampler_top_p {
/* .p = */ p ,
/* .min_keep = */ min_keep ,
} ,
} ;
}
// min-p
struct llama_sampler_min_p {
const float p ;
const size_t min_keep ;
} ;
static const char * llama_sampler_min_p_name ( const struct llama_sampler * /*smpl*/ ) {
return " min-p " ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
}
2024-10-17 18:59:52 +00:00
static void llama_sampler_min_p_apply ( struct llama_sampler * smpl , llama_token_data_array * cur_p ) {
const auto * ctx = ( llama_sampler_min_p * ) smpl - > ctx ;
if ( ctx - > p < = 0.0f | | ! cur_p - > size ) {
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
return ;
}
bool min_p_applied = false ;
2024-10-17 18:59:52 +00:00
// if the cur_p aren't sorted, try the unsorted implementation first
if ( ! cur_p - > sorted ) {
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
std : : vector < llama_token_data > filtered_tokens ;
float max_logit = - FLT_MAX ;
2024-10-17 18:59:52 +00:00
for ( size_t i = 0 ; i < cur_p - > size ; + + i ) {
max_logit = std : : max ( max_logit , cur_p - > data [ i ] . logit ) ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
}
2024-10-17 18:59:52 +00:00
const float min_logit = max_logit + logf ( ctx - > p ) ; // min logit for p_i >= p * p_max
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
2024-10-17 18:59:52 +00:00
for ( size_t i = 0 ; i < cur_p - > size ; + + i ) {
if ( cur_p - > data [ i ] . logit > = min_logit ) {
filtered_tokens . push_back ( cur_p - > data [ i ] ) ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
}
}
// if we have enough values the operation was a success
2024-10-17 18:59:52 +00:00
if ( filtered_tokens . size ( ) > = ctx - > min_keep ) {
memcpy ( cur_p - > data , filtered_tokens . data ( ) , filtered_tokens . size ( ) * sizeof ( llama_token_data ) ) ;
cur_p - > size = filtered_tokens . size ( ) ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
min_p_applied = true ;
}
}
2024-10-17 18:59:52 +00:00
// if the cur_p are sorted or the unsorted implementation failed, use this implementation
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
if ( ! min_p_applied ) {
// Sort the logits in descending order
2024-10-17 18:59:52 +00:00
if ( ! cur_p - > sorted ) {
std : : sort ( cur_p - > data , cur_p - > data + cur_p - > size , [ ] ( const llama_token_data & a , const llama_token_data & b ) {
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
return a . logit > b . logit ;
} ) ;
2024-10-17 18:59:52 +00:00
cur_p - > sorted = true ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
}
2024-10-17 18:59:52 +00:00
const float min_logit = cur_p - > data [ 0 ] . logit + logf ( ctx - > p ) ; // min logit for p_i >= p * p_max
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
size_t i = 1 ; // first token always matches
2024-10-17 18:59:52 +00:00
for ( ; i < cur_p - > size ; + + i ) {
if ( cur_p - > data [ i ] . logit < min_logit & & i > = ctx - > min_keep ) {
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
break ; // prob too small
}
}
// Resize the output vector to keep only the matching tokens
2024-10-17 18:59:52 +00:00
cur_p - > size = i ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
}
2024-10-17 18:59:52 +00:00
}
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
2024-10-17 18:59:52 +00:00
static struct llama_sampler * llama_sampler_min_p_clone ( const struct llama_sampler * smpl ) {
const auto * ctx = ( const llama_sampler_min_p * ) smpl - > ctx ;
return llama_sampler_init_min_p ( ctx - > p , ctx - > min_keep ) ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
}
2024-10-17 18:59:52 +00:00
static void llama_sampler_min_p_free ( struct llama_sampler * smpl ) {
delete ( llama_sampler_min_p * ) smpl - > ctx ;
}
static struct llama_sampler_i llama_sampler_min_p_i = {
/* .name = */ llama_sampler_min_p_name ,
/* .accept = */ nullptr ,
/* .apply = */ llama_sampler_min_p_apply ,
/* .reset = */ nullptr ,
/* .clone = */ llama_sampler_min_p_clone ,
/* .free = */ llama_sampler_min_p_free ,
} ;
struct llama_sampler * llama_sampler_init_min_p ( float p , size_t min_keep ) {
return new llama_sampler {
/* .iface = */ & llama_sampler_min_p_i ,
/* .ctx = */ new llama_sampler_min_p {
/* .p = */ p ,
/* .min_keep = */ min_keep ,
} ,
} ;
}
// tail-free
struct llama_sampler_tail_free {
const float z ;
const size_t min_keep ;
} ;
static const char * llama_sampler_tail_free_name ( const struct llama_sampler * /*smpl*/ ) {
return " tail-free " ;
}
static void llama_sampler_tail_free_apply ( struct llama_sampler * smpl , llama_token_data_array * cur_p ) {
const auto * ctx = ( llama_sampler_tail_free * ) smpl - > ctx ;
if ( ctx - > z > = 1.0f | | cur_p - > size < = 2 ) {
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
return ;
}
2024-10-17 18:59:52 +00:00
llama_sampler_softmax_impl ( cur_p ) ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
// Compute the first and second derivatives
2024-10-17 18:59:52 +00:00
std : : vector < float > first_derivatives ( cur_p - > size - 1 ) ;
std : : vector < float > second_derivatives ( cur_p - > size - 2 ) ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
for ( size_t i = 0 ; i < first_derivatives . size ( ) ; + + i ) {
2024-10-17 18:59:52 +00:00
first_derivatives [ i ] = cur_p - > data [ i ] . p - cur_p - > data [ i + 1 ] . p ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
}
for ( size_t i = 0 ; i < second_derivatives . size ( ) ; + + i ) {
second_derivatives [ i ] = first_derivatives [ i ] - first_derivatives [ i + 1 ] ;
}
// Calculate absolute value of second derivatives
for ( size_t i = 0 ; i < second_derivatives . size ( ) ; + + i ) {
second_derivatives [ i ] = std : : abs ( second_derivatives [ i ] ) ;
}
// Normalize the second derivatives
{
const float second_derivatives_sum = std : : accumulate ( second_derivatives . begin ( ) , second_derivatives . end ( ) , 0.0f ) ;
if ( second_derivatives_sum > 1e-6 f ) {
for ( float & value : second_derivatives ) {
value / = second_derivatives_sum ;
}
} else {
for ( float & value : second_derivatives ) {
value = 1.0f / second_derivatives . size ( ) ;
}
}
}
float cum_sum = 0.0f ;
2024-10-17 18:59:52 +00:00
size_t last_idx = cur_p - > size ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
for ( size_t i = 0 ; i < second_derivatives . size ( ) ; + + i ) {
cum_sum + = second_derivatives [ i ] ;
// Check if the running sum is greater than z or if we have kept at least min_keep tokens
2024-10-17 18:59:52 +00:00
if ( cum_sum > ctx - > z & & i > = ctx - > min_keep ) {
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
last_idx = i ;
break ;
}
}
// Resize the output vector to keep only the tokens above the tail location
2024-10-17 18:59:52 +00:00
cur_p - > size = last_idx ;
}
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
2024-10-17 18:59:52 +00:00
static struct llama_sampler * llama_sampler_tail_free_clone ( const struct llama_sampler * smpl ) {
const auto * ctx = ( const llama_sampler_tail_free * ) smpl - > ctx ;
return llama_sampler_init_tail_free ( ctx - > z , ctx - > min_keep ) ;
}
static void llama_sampler_tail_free_free ( struct llama_sampler * smpl ) {
delete ( llama_sampler_tail_free * ) smpl - > ctx ;
}
static struct llama_sampler_i llama_sampler_tail_free_i = {
/* .name = */ llama_sampler_tail_free_name ,
/* .accept = */ nullptr ,
/* .apply = */ llama_sampler_tail_free_apply ,
/* .reset = */ nullptr ,
/* .clone = */ llama_sampler_tail_free_clone ,
/* .free = */ llama_sampler_tail_free_free ,
} ;
struct llama_sampler * llama_sampler_init_tail_free ( float z , size_t min_keep ) {
return new llama_sampler {
/* .iface = */ & llama_sampler_tail_free_i ,
/* .ctx = */ new llama_sampler_tail_free {
/* .z = */ z ,
/*. min_keep = */ min_keep ,
} ,
} ;
}
// typical
struct llama_sampler_typical {
const float p ;
const size_t min_keep ;
} ;
static const char * llama_sampler_typical_name ( const struct llama_sampler * /*smpl*/ ) {
return " typical " ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
}
2024-10-17 18:59:52 +00:00
static void llama_sampler_typical_apply ( struct llama_sampler * smpl , llama_token_data_array * cur_p ) {
const auto * ctx = ( llama_sampler_typical * ) smpl - > ctx ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
// Reference implementation:
// https://github.com/huggingface/transformers/compare/main...cimeister:typical-sampling:typical-pr
2024-10-17 18:59:52 +00:00
if ( ctx - > p > = 1.0f ) {
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
return ;
}
// Compute the softmax of logits and calculate entropy
2024-10-17 18:59:52 +00:00
llama_sampler_softmax_impl ( cur_p ) ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
float entropy = 0.0f ;
2024-10-17 18:59:52 +00:00
for ( size_t i = 0 ; i < cur_p - > size ; + + i ) {
entropy + = - cur_p - > data [ i ] . p * logf ( cur_p - > data [ i ] . p ) ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
}
// Compute the absolute difference between negative log probability and entropy for each candidate
std : : vector < float > shifted_scores ;
2024-10-17 18:59:52 +00:00
for ( size_t i = 0 ; i < cur_p - > size ; + + i ) {
float shifted_score = fabsf ( - logf ( cur_p - > data [ i ] . p ) - entropy ) ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
shifted_scores . push_back ( shifted_score ) ;
}
// Sort tokens based on the shifted_scores and their corresponding indices
2024-10-17 18:59:52 +00:00
std : : vector < size_t > indices ( cur_p - > size ) ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
std : : iota ( indices . begin ( ) , indices . end ( ) , 0 ) ;
std : : sort ( indices . begin ( ) , indices . end ( ) , [ & ] ( size_t a , size_t b ) {
return shifted_scores [ a ] < shifted_scores [ b ] ;
} ) ;
// Compute the cumulative probabilities
float cum_sum = 0.0f ;
size_t last_idx = indices . size ( ) ;
for ( size_t i = 0 ; i < indices . size ( ) ; + + i ) {
size_t idx = indices [ i ] ;
2024-10-17 18:59:52 +00:00
cum_sum + = cur_p - > data [ idx ] . p ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
// Check if the running sum is greater than typical or if we have kept at least min_keep tokens
2024-10-17 18:59:52 +00:00
if ( cum_sum > ctx - > p & & i > = ctx - > min_keep - 1 ) {
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
last_idx = i + 1 ;
break ;
}
}
// Resize the output vector to keep only the locally typical tokens
2024-10-17 18:59:52 +00:00
std : : vector < llama_token_data > cur_p_new ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
for ( size_t i = 0 ; i < last_idx ; + + i ) {
size_t idx = indices [ i ] ;
2024-10-17 18:59:52 +00:00
cur_p_new . push_back ( cur_p - > data [ idx ] ) ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
}
2024-10-17 18:59:52 +00:00
// Replace the data in cur_p with the cur_p_new data
std : : copy ( cur_p_new . begin ( ) , cur_p_new . end ( ) , cur_p - > data ) ;
cur_p - > size = cur_p_new . size ( ) ;
cur_p - > sorted = false ;
}
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
2024-10-17 18:59:52 +00:00
static struct llama_sampler * llama_sampler_typical_clone ( const struct llama_sampler * smpl ) {
const auto * ctx = ( const llama_sampler_typical * ) smpl - > ctx ;
return llama_sampler_init_typical ( ctx - > p , ctx - > min_keep ) ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
}
2024-10-17 18:59:52 +00:00
static void llama_sampler_typical_free ( struct llama_sampler * smpl ) {
delete ( llama_sampler_typical * ) smpl - > ctx ;
}
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
2024-10-17 18:59:52 +00:00
static struct llama_sampler_i llama_sampler_typical_i = {
/* .name = */ llama_sampler_typical_name ,
/* .accept = */ nullptr ,
/* .apply = */ llama_sampler_typical_apply ,
/* .reset = */ nullptr ,
/* .clone = */ llama_sampler_typical_clone ,
/* .free = */ llama_sampler_typical_free ,
} ;
struct llama_sampler * llama_sampler_init_typical ( float p , size_t min_keep ) {
return new llama_sampler {
/* .iface = */ & llama_sampler_typical_i ,
/* .ctx = */ new llama_sampler_typical {
/* .p = */ p ,
/* .min_keep = */ min_keep ,
} ,
} ;
}
// temp
struct llama_sampler_temp {
const float temp ;
} ;
static const char * llama_sampler_temp_name ( const struct llama_sampler * /*smpl*/ ) {
return " temp " ;
}
static void llama_sampler_temp_apply ( struct llama_sampler * smpl , llama_token_data_array * cur_p ) {
const auto * ctx = ( llama_sampler_temp * ) smpl - > ctx ;
for ( size_t i = 0 ; i < cur_p - > size ; + + i ) {
cur_p - > data [ i ] . logit / = ctx - > temp ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
}
2024-10-17 18:59:52 +00:00
}
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
2024-10-17 18:59:52 +00:00
static struct llama_sampler * llama_sampler_temp_clone ( const struct llama_sampler * smpl ) {
const auto * ctx = ( const llama_sampler_temp * ) smpl - > ctx ;
return llama_sampler_init_temp ( ctx - > temp ) ;
}
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
2024-10-17 18:59:52 +00:00
static void llama_sampler_temp_free ( struct llama_sampler * smpl ) {
delete ( llama_sampler_temp * ) smpl - > ctx ;
}
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
2024-10-17 18:59:52 +00:00
static struct llama_sampler_i llama_sampler_temp_i = {
/* .name = */ llama_sampler_temp_name ,
/* .accept = */ nullptr ,
/* .apply = */ llama_sampler_temp_apply ,
/* .reset = */ nullptr ,
/* .clone = */ llama_sampler_temp_clone ,
/* .free = */ llama_sampler_temp_free ,
} ;
struct llama_sampler * llama_sampler_init_temp ( float temp ) {
return new llama_sampler {
/* .iface = */ & llama_sampler_temp_i ,
/* .ctx = */ new llama_sampler_temp {
/*.temp = */ temp ,
} ,
} ;
}
// temp-ext
struct llama_sampler_temp_ext {
const float temp ;
const float delta ;
const float exponent ;
} ;
static const char * llama_sampler_temp_ext_name ( const struct llama_sampler * /*smpl*/ ) {
return " temp-ext " ;
}
static void llama_sampler_temp_ext_apply ( struct llama_sampler * smpl , llama_token_data_array * cur_p ) {
const auto * ctx = ( llama_sampler_temp_ext * ) smpl - > ctx ;
if ( ctx - > delta > 0 ) {
const float min_temp = std : : max ( 0.0f , ctx - > temp - ctx - > delta ) ;
const float max_temp = ctx - > temp + ctx - > delta ;
float exponent_val = ctx - > exponent ;
// no need to do anything if there is only one (or zero) candidates
if ( cur_p - > size < = 1 ) {
return ;
}
// Calculate maximum possible entropy
float max_entropy = - logf ( 1.0f / cur_p - > size ) ;
llama_sampler_softmax_impl ( cur_p ) ;
// Calculate entropy of the softmax probabilities
float entropy = 0.0f ;
for ( size_t i = 0 ; i < cur_p - > size ; + + i ) {
float prob = cur_p - > data [ i ] . p ;
if ( prob > 0.0f ) { // Ensure no log(0)
entropy - = prob * logf ( prob ) ;
}
}
// Normalize the entropy (max_entropy cannot be 0 here because we checked cur_p->size != 1 above)
float normalized_entropy = entropy / max_entropy ;
// Map the normalized entropy to the desired temperature range using the power function
float dyn_temp = min_temp + ( max_temp - min_temp ) * powf ( normalized_entropy , exponent_val ) ;
# ifdef DEBUG
LLAMA_LOG_INFO ( " Your text maxtemp value is: %f \n " , max_temp ) ;
LLAMA_LOG_INFO ( " Entropy: %f \n " , entropy ) ;
LLAMA_LOG_INFO ( " Max Possible Entropy: %f \n " , max_entropy ) ;
LLAMA_LOG_INFO ( " Normalized Entropy: %f \n " , normalized_entropy ) ;
LLAMA_LOG_INFO ( " Exponent: %f \n " , exponent_val ) ;
LLAMA_LOG_INFO ( " Dynamic Temperature (dyn_temp): %f \n " , dyn_temp ) ;
# endif
// Apply the dynamically calculated temperature scaling
for ( size_t i = 0 ; i < cur_p - > size ; + + i ) {
cur_p - > data [ i ] . logit / = dyn_temp ;
}
// Re-compute softmax probabilities after scaling logits with dynamic temperature
const double max_l_double = cur_p - > data [ 0 ] . logit ;
double cum_sum_double = 0.0 ;
for ( size_t i = 0 ; i < cur_p - > size ; + + i ) {
double p = exp ( cur_p - > data [ i ] . logit - max_l_double ) ;
cur_p - > data [ i ] . p = p ; // Store the scaled probability
cum_sum_double + = p ;
}
for ( size_t i = 0 ; i < cur_p - > size ; + + i ) {
cur_p - > data [ i ] . p / = cum_sum_double ; // Re-normalize the probabilities
}
# ifdef DEBUG
// Print the updated top 25 probabilities after temperature scaling
LLAMA_LOG_INFO ( " \n Updated Top 25 Probabilities After Dynamic Temperature Scaling (in percentages): \n " ) ;
for ( size_t i = 0 ; i < 25 & & i < cur_p - > size ; + + i ) {
LLAMA_LOG_INFO ( " Token %zu: %f%% \n " , i + 1 , cur_p - > data [ i ] . p * 100.0f ) ;
}
# endif
} else {
for ( size_t i = 0 ; i < cur_p - > size ; + + i ) {
cur_p - > data [ i ] . logit / = ctx - > temp ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
}
}
2024-10-17 18:59:52 +00:00
}
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
2024-10-17 18:59:52 +00:00
static struct llama_sampler * llama_sampler_temp_ext_clone ( const struct llama_sampler * smpl ) {
const auto * ctx = ( const llama_sampler_temp_ext * ) smpl - > ctx ;
return llama_sampler_init_temp_ext ( ctx - > temp , ctx - > delta , ctx - > exponent ) ;
}
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
2024-10-17 18:59:52 +00:00
static void llama_sampler_temp_ext_free ( struct llama_sampler * smpl ) {
delete ( llama_sampler_temp_ext * ) smpl - > ctx ;
}
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
2024-10-17 18:59:52 +00:00
static struct llama_sampler_i llama_sampler_temp_ext_i = {
/* .name = */ llama_sampler_temp_ext_name ,
/* .accept = */ nullptr ,
/* .apply = */ llama_sampler_temp_ext_apply ,
/* .reset = */ nullptr ,
/* .clone = */ llama_sampler_temp_ext_clone ,
/* .free = */ llama_sampler_temp_ext_free ,
} ;
struct llama_sampler * llama_sampler_init_temp_ext ( float temp , float delta , float exponent ) {
return new llama_sampler {
/* .iface = */ & llama_sampler_temp_ext_i ,
/* .ctx = */ new llama_sampler_temp_ext {
/* .temp = */ temp ,
/* .delta = */ delta ,
/* .exponent = */ exponent ,
} ,
} ;
}
// mirostat
struct llama_sampler_mirostat {
const int32_t n_vocab ;
const uint32_t seed ;
uint32_t seed_cur ;
const float tau ;
const float eta ;
const int32_t m ;
float mu ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
2024-10-17 18:59:52 +00:00
std : : mt19937 rng ;
} ;
static const char * llama_sampler_mirostat_name ( const struct llama_sampler * /*smpl*/ ) {
return " mirostat " ;
}
static void llama_sampler_mirostat_apply ( struct llama_sampler * smpl , llama_token_data_array * cur_p ) {
auto * ctx = ( llama_sampler_mirostat * ) smpl - > ctx ;
llama_sampler_softmax_impl ( cur_p ) ;
// Estimate s_hat using the most probable m tokens
float s_hat = 0.0 ;
float sum_ti_bi = 0.0 ;
float sum_ti_sq = 0.0 ;
for ( size_t i = 0 ; i < size_t ( ctx - > m - 1 ) & & i < cur_p - > size - 1 ; + + i ) {
float t_i = logf ( float ( i + 2 ) / float ( i + 1 ) ) ;
float b_i = logf ( cur_p - > data [ i ] . p / cur_p - > data [ i + 1 ] . p ) ;
sum_ti_bi + = t_i * b_i ;
sum_ti_sq + = t_i * t_i ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
}
2024-10-17 18:59:52 +00:00
s_hat = sum_ti_bi / sum_ti_sq ;
// Compute k from the estimated s_hat and target surprise value
float epsilon_hat = s_hat - 1 ;
float k = powf ( ( epsilon_hat * powf ( 2 , ctx - > mu ) ) / ( 1 - powf ( ctx - > n_vocab , - epsilon_hat ) ) , 1 / s_hat ) ;
llama_sampler_top_k_impl ( cur_p , std : : max ( int ( k ) , 1 ) ) ;
llama_sampler_softmax_impl ( cur_p ) ;
const int idx = llama_sample_dist ( cur_p , ctx - > rng ) ;
cur_p - > selected = idx ;
float observed_surprise = - log2f ( cur_p - > data [ idx ] . p ) ;
float e = observed_surprise - ctx - > tau ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
2024-10-17 18:59:52 +00:00
// Update mu using the learning rate and error
ctx - > mu = ctx - > mu - ctx - > eta * e ;
}
static struct llama_sampler * llama_sampler_mirostat_clone ( const struct llama_sampler * smpl ) {
const auto * ctx = ( const llama_sampler_mirostat * ) smpl - > ctx ;
auto * result = llama_sampler_init_mirostat ( ctx - > n_vocab , ctx - > seed , ctx - > tau , ctx - > eta , ctx - > m ) ;
// copy the state
{
auto * result_ctx = ( llama_sampler_mirostat * ) smpl - > ctx ;
result_ctx - > mu = ctx - > mu ;
result_ctx - > rng = ctx - > rng ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
}
2024-10-17 18:59:52 +00:00
return result ;
}
static void llama_sampler_mirostat_reset ( struct llama_sampler * smpl ) {
auto * ctx = ( llama_sampler_mirostat * ) smpl - > ctx ;
ctx - > mu = 2.0f * ctx - > tau ;
ctx - > seed_cur = get_rng_seed ( ctx - > seed ) ;
ctx - > rng . seed ( ctx - > seed_cur ) ;
}
static void llama_sampler_mirostat_free ( struct llama_sampler * smpl ) {
delete ( llama_sampler_mirostat * ) smpl - > ctx ;
}
static struct llama_sampler_i llama_sampler_mirostat_i = {
/* .name = */ llama_sampler_mirostat_name ,
/* .accept = */ nullptr ,
/* .apply = */ llama_sampler_mirostat_apply ,
/* .reset = */ llama_sampler_mirostat_reset ,
/* .clone = */ llama_sampler_mirostat_clone ,
/* .free = */ llama_sampler_mirostat_free ,
} ;
struct llama_sampler * llama_sampler_init_mirostat ( int32_t n_vocab , uint32_t seed , float tau , float eta , int32_t m ) {
auto seed_cur = get_rng_seed ( seed ) ;
return new llama_sampler {
/* .iface = */ & llama_sampler_mirostat_i ,
/* .ctx = */ new llama_sampler_mirostat {
/* .n_vocab = */ n_vocab ,
/* .seed = */ seed ,
/* .seed_cur = */ seed_cur ,
/* .tau = */ tau ,
/* .eta = */ eta ,
/* .m = */ m ,
/* .mu = */ 2.0f * tau ,
/* .rng = */ std : : mt19937 ( seed_cur ) ,
} ,
} ;
}
// mirostat v2
struct llama_sampler_mirostat_v2 {
const uint32_t seed ;
uint32_t seed_cur ;
const float tau ;
const float eta ;
float mu ;
std : : mt19937 rng ;
} ;
static const char * llama_sampler_mirostat_v2_name ( const struct llama_sampler * /*smpl*/ ) {
return " mirostat-v2 " ;
}
static void llama_sampler_mirostat_v2_apply ( struct llama_sampler * smpl , llama_token_data_array * cur_p ) {
auto * ctx = ( llama_sampler_mirostat_v2 * ) smpl - > ctx ;
llama_sampler_softmax_impl ( cur_p ) ;
// Truncate the words with surprise values greater than mu
cur_p - > size = std : : distance ( cur_p - > data , std : : find_if ( cur_p - > data , cur_p - > data + cur_p - > size , [ & ] ( const llama_token_data & candidate ) {
return - log2f ( candidate . p ) > ctx - > mu ;
} ) ) ;
if ( cur_p - > size = = 0 ) {
cur_p - > size = 1 ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
}
2024-10-17 18:59:52 +00:00
// Normalize the probabilities of the remaining words
llama_sampler_softmax_impl ( cur_p ) ;
const int idx = llama_sample_dist ( cur_p , ctx - > rng ) ;
cur_p - > selected = idx ;
float observed_surprise = - log2f ( cur_p - > data [ idx ] . p ) ;
float e = observed_surprise - ctx - > tau ;
// Update mu using the learning rate and error
ctx - > mu = ctx - > mu - ctx - > eta * e ;
}
static void llama_sampler_mirostat_v2_reset ( struct llama_sampler * smpl ) {
auto * ctx = ( llama_sampler_mirostat_v2 * ) smpl - > ctx ;
ctx - > mu = 2.0f * ctx - > tau ;
ctx - > seed_cur = get_rng_seed ( ctx - > seed ) ;
ctx - > rng . seed ( ctx - > seed_cur ) ;
}
static struct llama_sampler * llama_sampler_mirostat_v2_clone ( const struct llama_sampler * smpl ) {
const auto * ctx = ( const llama_sampler_mirostat_v2 * ) smpl - > ctx ;
auto * result = llama_sampler_init_mirostat_v2 ( ctx - > seed , ctx - > tau , ctx - > eta ) ;
// copy the state
{
auto * result_ctx = ( llama_sampler_mirostat_v2 * ) result - > ctx ;
result_ctx - > mu = ctx - > mu ;
result_ctx - > rng = ctx - > rng ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
}
2024-10-17 18:59:52 +00:00
return result ;
}
static void llama_sampler_mirostat_v2_free ( struct llama_sampler * smpl ) {
delete ( llama_sampler_mirostat_v2 * ) smpl - > ctx ;
}
static struct llama_sampler_i llama_sampler_mirostat_v2_i = {
/* .name = */ llama_sampler_mirostat_v2_name ,
/* .accept = */ nullptr ,
/* .apply = */ llama_sampler_mirostat_v2_apply ,
/* .reset = */ llama_sampler_mirostat_v2_reset ,
/* .clone = */ llama_sampler_mirostat_v2_clone ,
/* .free = */ llama_sampler_mirostat_v2_free ,
} ;
struct llama_sampler * llama_sampler_init_mirostat_v2 ( uint32_t seed , float tau , float eta ) {
auto seed_cur = get_rng_seed ( seed ) ;
return new llama_sampler {
/* .iface = */ & llama_sampler_mirostat_v2_i ,
/* .ctx = */ new llama_sampler_mirostat_v2 {
/* .seed = */ seed ,
/* .seed_cur = */ seed_cur ,
/* .tau = */ tau ,
/* .eta = */ eta ,
/* .mu = */ 2.0f * tau ,
/* .rng = */ std : : mt19937 ( seed_cur ) ,
} ,
} ;
}
// grammar
struct llama_sampler_grammar {
const struct llama_vocab * vocab ;
std : : string grammar_str ;
std : : string grammar_root ;
struct llama_grammar * grammar ;
} ;
static const char * llama_sampler_grammar_name ( const struct llama_sampler * /*smpl*/ ) {
return " grammar " ;
}
static void llama_sampler_grammar_accept_impl ( struct llama_sampler * smpl , llama_token token ) {
auto * ctx = ( llama_sampler_grammar * ) smpl - > ctx ;
if ( ctx - > grammar ) {
llama_grammar_accept_impl ( * ctx - > grammar , token ) ;
}
}
static void llama_sampler_grammar_apply ( struct llama_sampler * smpl , llama_token_data_array * cur_p ) {
auto * ctx = ( llama_sampler_grammar * ) smpl - > ctx ;
if ( ctx - > grammar ) {
llama_grammar_apply_impl ( * ctx - > grammar , cur_p ) ;
}
}
static void llama_sampler_grammar_reset ( struct llama_sampler * smpl ) {
auto * ctx = ( llama_sampler_grammar * ) smpl - > ctx ;
if ( ! ctx - > grammar ) {
return ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
}
2024-10-17 18:59:52 +00:00
auto * grammar_new = llama_grammar_init_impl ( ctx - > grammar - > vocab , ctx - > grammar_str . c_str ( ) , ctx - > grammar_root . c_str ( ) ) ;
llama_grammar_free_impl ( ctx - > grammar ) ;
ctx - > grammar = grammar_new ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
}
2024-10-17 18:59:52 +00:00
static struct llama_sampler * llama_sampler_grammar_clone ( const struct llama_sampler * smpl ) {
const auto * ctx = ( const llama_sampler_grammar * ) smpl - > ctx ;
auto * result = llama_sampler_init_grammar_impl ( * ctx - > vocab , nullptr , nullptr ) ;
// copy the state
{
auto * result_ctx = ( llama_sampler_grammar * ) result - > ctx ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
2024-10-17 18:59:52 +00:00
if ( ctx - > grammar ) {
result_ctx - > grammar_str = ctx - > grammar_str ;
result_ctx - > grammar_root = ctx - > grammar_root ;
result_ctx - > grammar = llama_grammar_clone_impl ( * ctx - > grammar ) ;
}
}
return result ;
}
static void llama_sampler_grammar_free ( struct llama_sampler * smpl ) {
const auto * ctx = ( llama_sampler_grammar * ) smpl - > ctx ;
if ( ctx - > grammar ) {
llama_grammar_free_impl ( ctx - > grammar ) ;
}
delete ctx ;
}
static struct llama_sampler_i llama_sampler_grammar_i = {
/* .name = */ llama_sampler_grammar_name ,
/* .accept = */ llama_sampler_grammar_accept_impl ,
/* .apply = */ llama_sampler_grammar_apply ,
/* .reset = */ llama_sampler_grammar_reset ,
/* .clone = */ llama_sampler_grammar_clone ,
/* .free = */ llama_sampler_grammar_free ,
} ;
struct llama_sampler * llama_sampler_init_grammar_impl ( const struct llama_vocab & vocab , const char * grammar_str , const char * grammar_root ) {
auto * ctx = new llama_sampler_grammar ;
if ( grammar_str ! = nullptr & & grammar_str [ 0 ] ! = ' \0 ' ) {
* ctx = {
/* .vocab = */ & vocab ,
/* .grammar_str = */ grammar_str ,
/* .grammar_root = */ grammar_root ,
/* .grammar = */ llama_grammar_init_impl ( & vocab , grammar_str , grammar_root ) ,
} ;
} else {
* ctx = {
/* .vocab = */ & vocab ,
/* .grammar_str = */ { } ,
/* .grammar_root = */ { } ,
/* .grammar = */ nullptr ,
} ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
}
2024-10-17 18:59:52 +00:00
return new llama_sampler {
/* .iface = */ & llama_sampler_grammar_i ,
/* .ctx = */ ctx ,
} ;
}
// penalties
struct llama_sampler_penalties {
const int32_t n_vocab ;
const llama_token special_eos_id ;
const llama_token linefeed_id ;
const int32_t penalty_last_n ;
const float penalty_repeat ;
const float penalty_freq ;
const float penalty_present ;
const bool penalize_nl ;
const bool ignore_eos ;
ring_buffer < llama_token > prev ;
} ;
static const char * llama_sampler_penalties_name ( const struct llama_sampler * /*smpl*/ ) {
return " penalties " ;
}
static void llama_sampler_penalties_accept ( struct llama_sampler * smpl , llama_token token ) {
auto * ctx = ( llama_sampler_penalties * ) smpl - > ctx ;
if ( ctx - > penalty_last_n = = 0 ) {
return ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
}
2024-10-17 18:59:52 +00:00
ctx - > prev . push_back ( token ) ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
}
2024-10-17 18:59:52 +00:00
static void llama_sampler_penalties_apply ( struct llama_sampler * smpl , llama_token_data_array * cur_p ) {
auto * ctx = ( llama_sampler_penalties * ) smpl - > ctx ;
if ( ctx - > ignore_eos ) {
assert ( ctx - > special_eos_id > = 0 ) ;
// optimistically check if the candidates are not yet sorted/shuffled/truncated
if ( cur_p - > size > ( size_t ) ctx - > special_eos_id & & cur_p - > data [ ctx - > special_eos_id ] . id = = ctx - > special_eos_id ) {
cur_p - > data [ ctx - > special_eos_id ] . logit = - INFINITY ;
} else {
// else, search for the special EOS token
for ( size_t i = 0 ; i < cur_p - > size ; + + i ) {
if ( cur_p - > data [ i ] . id = = ctx - > special_eos_id ) {
cur_p - > data [ i ] . logit = - INFINITY ;
break ;
}
}
}
}
if ( ( ctx - > penalty_last_n = = 0 ) | |
( ctx - > penalty_repeat = = 1.0f & & ctx - > penalty_freq = = 0.0f & & ctx - > penalty_present = = 0.0f ) ) {
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
return ;
}
2024-10-17 18:59:52 +00:00
bool nl_found = false ;
size_t nl_idx = 0 ;
float nl_logit = - INFINITY ;
if ( ! ctx - > penalize_nl ) {
assert ( ctx - > linefeed_id > = 0 ) ;
// optimistically check if the candidates are not yet sorted/shuffled/truncated
if ( cur_p - > size > ( size_t ) ctx - > linefeed_id & & cur_p - > data [ ctx - > linefeed_id ] . id = = ctx - > linefeed_id ) {
nl_found = true ;
nl_idx = ctx - > linefeed_id ;
nl_logit = cur_p - > data [ ctx - > linefeed_id ] . logit ;
} else {
// else, search for the linefeed token
for ( size_t i = 0 ; i < cur_p - > size ; + + i ) {
if ( cur_p - > data [ i ] . id = = ctx - > linefeed_id ) {
nl_found = true ;
nl_idx = i ;
nl_logit = cur_p - > data [ i ] . logit ;
break ;
}
}
}
}
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
// Create a frequency map to count occurrences of each token in last_tokens
2024-10-17 18:59:52 +00:00
// TODO: optimize this by maintaining the token count in the sampler context
using llama_token_cnt = std : : unordered_map < llama_token , int > ;
llama_token_cnt token_count ;
for ( int i = 0 ; i < std : : min < int > ( ctx - > penalty_last_n , ctx - > prev . size ( ) ) ; + + i ) {
token_count [ ctx - > prev . rat ( i ) ] + + ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
}
2024-10-17 18:59:52 +00:00
// Apply frequency and presence penalties to the cur_p
for ( size_t i = 0 ; i < cur_p - > size ; + + i ) {
const auto token_iter = token_count . find ( cur_p - > data [ i ] . id ) ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
if ( token_iter = = token_count . end ( ) ) {
continue ;
}
const int count = token_iter - > second ;
// The academic publication that described this technique actually just only divided, but that would cause tokens with negative logits to become more likely, which is obviously wrong.
// This is common fix for this problem, which is to multiply by the penalty instead of dividing.
2024-10-17 18:59:52 +00:00
if ( cur_p - > data [ i ] . logit < = 0 ) {
cur_p - > data [ i ] . logit * = ctx - > penalty_repeat ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
} else {
2024-10-17 18:59:52 +00:00
cur_p - > data [ i ] . logit / = ctx - > penalty_repeat ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
}
2024-10-17 18:59:52 +00:00
cur_p - > data [ i ] . logit - = float ( count ) * ctx - > penalty_freq + float ( count > 0 ) * ctx - > penalty_present ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
}
2024-10-17 18:59:52 +00:00
cur_p - > sorted = false ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
2024-10-17 18:59:52 +00:00
if ( ! ctx - > penalize_nl & & nl_found ) {
// restore the logit of the newline token if it was penalized
cur_p - > data [ nl_idx ] . logit = nl_logit ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
}
}
2024-10-17 18:59:52 +00:00
static void llama_sampler_penalties_reset ( struct llama_sampler * smpl ) {
auto * ctx = ( llama_sampler_penalties * ) smpl - > ctx ;
ctx - > prev . clear ( ) ;
}
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
2024-10-17 18:59:52 +00:00
static struct llama_sampler * llama_sampler_penalties_clone ( const struct llama_sampler * smpl ) {
const auto * ctx = ( const llama_sampler_penalties * ) smpl - > ctx ;
auto * result = llama_sampler_init_penalties (
ctx - > n_vocab ,
ctx - > special_eos_id ,
ctx - > linefeed_id ,
ctx - > penalty_last_n ,
ctx - > penalty_repeat ,
ctx - > penalty_freq ,
ctx - > penalty_present ,
ctx - > penalize_nl ,
ctx - > ignore_eos ) ;
// copy the state
{
auto * result_ctx = ( llama_sampler_penalties * ) result - > ctx ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
2024-10-17 18:59:52 +00:00
result_ctx - > prev = ctx - > prev ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
}
2024-10-17 18:59:52 +00:00
return result ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
}
2024-10-17 18:59:52 +00:00
static void llama_sampler_penalties_free ( struct llama_sampler * smpl ) {
delete ( llama_sampler_penalties * ) smpl - > ctx ;
}
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
2024-10-17 18:59:52 +00:00
static struct llama_sampler_i llama_sampler_penalties_i = {
/* .name = */ llama_sampler_penalties_name ,
/* .accept = */ llama_sampler_penalties_accept ,
/* .apply = */ llama_sampler_penalties_apply ,
/* .reset = */ llama_sampler_penalties_reset ,
/* .clone = */ llama_sampler_penalties_clone ,
/* .free = */ llama_sampler_penalties_free ,
} ;
struct llama_sampler * llama_sampler_init_penalties (
int32_t n_vocab ,
llama_token special_eos_id ,
llama_token linefeed_id ,
int32_t penalty_last_n ,
float penalty_repeat ,
float penalty_freq ,
float penalty_present ,
bool penalize_nl ,
bool ignore_eos ) {
if ( linefeed_id = = LLAMA_TOKEN_NULL ) {
penalize_nl = true ;
}
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
2024-10-17 18:59:52 +00:00
if ( special_eos_id = = LLAMA_TOKEN_NULL ) {
ignore_eos = false ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
}
2024-10-17 18:59:52 +00:00
penalty_last_n = std : : max ( penalty_last_n , 0 ) ;
return new llama_sampler {
/* .iface = */ & llama_sampler_penalties_i ,
/* .ctx = */ new llama_sampler_penalties {
/* .n_vocab = */ n_vocab ,
/* .special_eos_id = */ special_eos_id ,
/* .linefeed_id = */ linefeed_id ,
/* .penalty_last_n = */ penalty_last_n ,
/* .penalty_repeat = */ penalty_repeat ,
/* .penalty_freq = */ penalty_freq ,
/* .penalty_present = */ penalty_present ,
/* .penalize_nl = */ penalize_nl ,
/* .ignore_eos = */ ignore_eos ,
/* .prev = */ ring_buffer < llama_token > ( penalty_last_n ) ,
} ,
} ;
}
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
2024-10-17 18:59:52 +00:00
// logit-bias
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
2024-10-17 18:59:52 +00:00
struct llama_sampler_logit_bias {
const int32_t n_vocab ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
2024-10-17 18:59:52 +00:00
const std : : vector < llama_logit_bias > logit_bias ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
2024-10-17 18:59:52 +00:00
std : : vector < llama_logit_bias > to_search ;
} ;
static const char * llama_sampler_logit_bias_name ( const struct llama_sampler * /*smpl*/ ) {
return " logit-bias " ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
}
2024-10-17 18:59:52 +00:00
static void llama_sampler_logit_bias_apply ( struct llama_sampler * smpl , llama_token_data_array * cur_p ) {
auto * ctx = ( llama_sampler_logit_bias * ) smpl - > ctx ;
if ( ctx - > logit_bias . empty ( ) ) {
return ;
}
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
2024-10-17 18:59:52 +00:00
ctx - > to_search . clear ( ) ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
2024-10-17 18:59:52 +00:00
// update the candidates that have not been shuffled in the vocabulary (i.e. idx == id)
for ( const auto & lb : ctx - > logit_bias ) {
if ( lb . token > = 0 & & cur_p - > size > ( size_t ) lb . token & & cur_p - > data [ lb . token ] . id = = lb . token ) {
cur_p - > data [ lb . token ] . logit + = lb . bias ;
} else {
ctx - > to_search . push_back ( lb ) ;
}
}
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
2024-10-17 18:59:52 +00:00
if ( ctx - > to_search . empty ( ) ) {
return ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
}
2024-10-17 18:59:52 +00:00
// search for the remaining candidates that were not found in the previous step
for ( size_t i = 0 ; i < cur_p - > size ; + + i ) {
for ( const auto & lb : ctx - > to_search ) {
if ( cur_p - > data [ i ] . id = = lb . token ) {
cur_p - > data [ i ] . logit + = lb . bias ;
break ;
}
}
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
}
2024-10-17 18:59:52 +00:00
}
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
2024-10-17 18:59:52 +00:00
static struct llama_sampler * llama_sampler_logit_bias_clone ( const struct llama_sampler * smpl ) {
const auto * ctx = ( const llama_sampler_logit_bias * ) smpl - > ctx ;
return llama_sampler_init_logit_bias ( ctx - > n_vocab , ctx - > logit_bias . size ( ) , ctx - > logit_bias . data ( ) ) ;
}
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
2024-10-17 18:59:52 +00:00
static void llama_sampler_logit_bias_free ( struct llama_sampler * smpl ) {
delete ( llama_sampler_logit_bias * ) smpl - > ctx ;
}
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
2024-10-17 18:59:52 +00:00
static struct llama_sampler_i llama_sampler_logit_bias_i = {
/* .name = */ llama_sampler_logit_bias_name ,
/* .accept = */ nullptr ,
/* .apply = */ llama_sampler_logit_bias_apply ,
/* .reset = */ nullptr ,
/* .clone = */ llama_sampler_logit_bias_clone ,
/* .free = */ llama_sampler_logit_bias_free ,
} ;
struct llama_sampler * llama_sampler_init_logit_bias (
int32_t n_vocab ,
int32_t n_logit_bias ,
const llama_logit_bias * logit_bias ) {
return new llama_sampler {
/* .iface = */ & llama_sampler_logit_bias_i ,
/* .ctx = */ new llama_sampler_logit_bias {
/* .n_vocab = */ n_vocab ,
/* .logit_bias = */ std : : vector < llama_logit_bias > ( logit_bias , logit_bias + n_logit_bias ) ,
/* .to_search = */ { } ,
} ,
} ;
}
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
2024-10-17 18:59:52 +00:00
// utils
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
2024-10-17 18:59:52 +00:00
uint32_t llama_sampler_get_seed ( const struct llama_sampler * smpl ) {
if ( smpl - > iface = = & llama_sampler_dist_i ) {
return ( ( const llama_sampler_dist * ) smpl - > ctx ) - > seed_cur ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
}
2024-10-17 18:59:52 +00:00
if ( smpl - > iface = = & llama_sampler_mirostat_i ) {
return ( ( const llama_sampler_mirostat * ) smpl - > ctx ) - > seed_cur ;
}
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
2024-10-17 18:59:52 +00:00
if ( smpl - > iface = = & llama_sampler_mirostat_v2_i ) {
return ( ( const llama_sampler_mirostat_v2 * ) smpl - > ctx ) - > seed_cur ;
}
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
2024-10-17 18:59:52 +00:00
if ( smpl - > iface = = & llama_sampler_chain_i ) {
const auto * ctx = ( const llama_sampler_chain * ) smpl - > ctx ;
for ( auto it = ctx - > samplers . rbegin ( ) ; it ! = ctx - > samplers . rend ( ) ; + + it ) {
const uint32_t seed = llama_sampler_get_seed ( * it ) ;
if ( seed ! = LLAMA_DEFAULT_SEED ) {
return seed ;
}
}
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
}
2024-10-17 18:59:52 +00:00
return LLAMA_DEFAULT_SEED ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
}
2024-10-17 18:59:52 +00:00
// perf
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
2024-10-17 18:59:52 +00:00
struct llama_perf_sampler_data llama_perf_sampler ( const struct llama_sampler * chain ) {
struct llama_perf_sampler_data data = { } ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
2024-10-17 18:59:52 +00:00
if ( chain = = nullptr | | chain - > iface ! = & llama_sampler_chain_i ) {
GGML_ABORT ( " %s: invalid sampler passed - requires a sampler created with llama_sampler_chain_init() \n " , __func__ ) ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
}
2024-10-17 18:59:52 +00:00
const auto * ctx = ( const struct llama_sampler_chain * ) chain - > ctx ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
2024-10-17 18:59:52 +00:00
data . t_sample_ms = 1e-3 * ctx - > t_sample_us ;
data . n_sample = std : : max ( 0 , ctx - > n_sample ) ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
2024-10-17 18:59:52 +00:00
return data ;
}
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
2024-10-17 18:59:52 +00:00
void llama_perf_sampler_print ( const struct llama_sampler * chain ) {
const auto data = llama_perf_sampler ( chain ) ;
LLAMA_LOG_INFO ( " %s: sampling time = %10.2f ms / %5d runs (%8.2f ms per token, %8.2f tokens per second) \n " ,
__func__ , data . t_sample_ms , data . n_sample , data . t_sample_ms / data . n_sample , 1e3 / data . t_sample_ms * data . n_sample ) ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
}
2024-10-17 18:59:52 +00:00
void llama_perf_sampler_reset ( struct llama_sampler * chain ) {
if ( chain = = nullptr | | chain - > iface ! = & llama_sampler_chain_i ) {
GGML_ABORT ( " %s: invalid sampler passed - requires a sampler created with llama_sampler_chain_init() \n " , __func__ ) ;
}
auto * ctx = ( struct llama_sampler_chain * ) chain - > ctx ;
ctx - > t_sample_us = ctx - > n_sample = 0 ;
Re-introduce the `llama` package (#5034)
* Re-introduce the llama package
This PR brings back the llama package, making it possible to call llama.cpp and
ggml APIs from Go directly via CGo. This has a few advantages:
- C APIs can be called directly from Go without needing to use the previous
"server" REST API
- On macOS and for CPU builds on Linux and Windows, Ollama can be built without
a go generate ./... step, making it easy to get up and running to hack on
parts of Ollama that don't require fast inference
- Faster build times for AVX,AVX2,CUDA and ROCM (a full build of all runners
takes <5 min on a fast CPU)
- No git submodule making it easier to clone and build from source
This is a big PR, but much of it is vendor code except for:
- llama.go CGo bindings
- example/: a simple example of running inference
- runner/: a subprocess server designed to replace the llm/ext_server package
- Makefile an as minimal as possible Makefile to build the runner package for
different targets (cpu, avx, avx2, cuda, rocm)
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
* cache: Clear old KV cache entries when evicting a slot
When forking a cache entry, if no empty slots are available we
evict the least recently used one and copy over the KV entries
from the closest match. However, this copy does not overwrite
existing values but only adds new ones. Therefore, we need to
clear the old slot first.
This change fixes two issues:
- The KV cache fills up and runs out of space even though we think
we are managing it correctly
- Performance gets worse over time as we use new cache entries that
are not hot in the processor caches
* doc: explain golang objc linker warning (#6830)
* llama: gather transitive dependencies for rocm for dist packaging (#6848)
* Refine go server makefiles to be more DRY (#6924)
This breaks up the monolithic Makefile for the Go based runners into a
set of utility files as well as recursive Makefiles for the runners.
Files starting with the name "Makefile" are buildable, while files that
end with ".make" are utilities to include in other Makefiles. This
reduces the amount of nearly identical targets and helps set a pattern
for future community contributions for new GPU runner architectures.
When we are ready to switch over to the Go runners, these files should
move to the top of the repo, and we should add targets for the main CLI,
as well as a helper "install" (put all the built binaries on the local
system in a runnable state) and "dist" target (generate the various
tar/zip files for distribution) for local developer use.
* llama: don't create extraneous directories (#6988)
* llama: Exercise the new build in CI (#6989)
Wire up some basic sanity testing in CI for the Go runner. GPU runners are not covered yet.
* llama: Refine developer docs for Go server (#6842)
This enhances the documentation for development focusing on the new Go
server. After we complete the transition further doc refinements
can remove the "transition" discussion.
* runner.go: Allocate batches for all sequences during init
We should tell the model that we could have full batches for all
sequences. We already do this when we allocate the batches but it was
missed during initialization.
* llama.go: Don't return nil from Tokenize on zero length input
Potentially receiving nil in a non-error condition is surprising to
most callers - it's better to return an empty slice.
* runner.go: Remove stop tokens from cache
If the last token is EOG then we don't return this and it isn't
present in the cache (because it was never submitted to Decode).
This works well for extending the cache entry with a new sequence.
However, for multi-token stop sequences, we won't return any of the
tokens but all but the last one will be in the cache. This means
when the conversation continues the cache will contain tokens that
don't overlap with the new prompt.
This works (we will pick up the portion where there is overlap) but
it causes unnecessary cache thrashing because we will fork the original
cache entry as it is not a perfect match.
By trimming the cache to the tokens that we actually return this
issue can be avoided.
* runner.go: Simplify flushing of pending tokens
* runner.go: Update TODOs
* runner.go: Don't panic when processing sequences
If there is an error processing a sequence, we should return a
clean HTTP error back to Ollama rather than panicing. This will
make us more resilient to transient failures.
Panics can still occur during startup as there is no way to serve
requests if that fails.
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: More accurately capture timings
Currently prompt processing time doesn't capture the that it takes
to tokenize the input, only decoding time. We should capture the
full process to more accurately reflect reality. This is especially
true once we start processing images where the initial processing
can take significant time. This is also more consistent with the
existing C++ runner.
* runner.go: Support for vision models
In addition to bringing feature parity with the C++ runner, this also
incorporates several improvements:
- Cache prompting works with images, avoiding the need to re-decode
embeddings for every message in a conversation
- Parallelism is supported, avoiding the need to restrict to one
sequence at a time. (Though for now Ollama will not schedule
them while we might need to fall back to the old runner.)
Co-authored-by: jmorganca <jmorganca@gmail.com>
* runner.go: Move Unicode checking code and add tests
* runner.go: Export external cache members
Runner and cache are in the same package so the change doesn't
affect anything but it is more internally consistent.
* runner.go: Image embedding cache
Generating embeddings from images can take significant time (on
my machine between 100ms and 8s depending on the model). Although
we already cache the result of decoding these images, the embeddings
need to be regenerated every time. This is not necessary if we get
the same image over and over again, for example, during a conversation.
This currently uses a very small cache with a very simple algorithm
but it is easy to improve as is warranted.
* llama: catch up on patches
Carry forward solar-pro and cli-unicode patches
* runner.go: Don't re-allocate memory for every batch
We can reuse memory allocated from batch to batch since batch
size is fixed. This both saves the cost of reallocation as well
keeps the cache lines hot.
This results in a roughly 1% performance improvement for token
generation with Nvidia GPUs on Linux.
* runner.go: Default to classic input cache policy
The input cache as part of the go runner implemented a cache
policy that aims to maximize hit rate in both single and multi-
user scenarios. When there is a cache hit, the response is
very fast.
However, performance is actually slower when there is an input
cache miss due to worse GPU VRAM locality. This means that
performance is generally better overall for multi-user scenarios
(better input cache hit rate, locality was relatively poor already).
But worse for single users (input cache hit rate is about the same,
locality is now worse).
This defaults the policy back to the old one to avoid a regression
but keeps the new one available through an environment variable
OLLAMA_MULTIUSER_CACHE. This is left undocumented as the goal is
to improve this in the future to get the best of both worlds
without user configuration.
For inputs that result in cache misses, on Nvidia/Linux this
change improves performance by 31% for prompt processing and
13% for token generation.
* runner.go: Increase size of response channel
Generally the CPU can easily keep up with handling reponses that
are generated but there's no reason not to let generation continue
and handle things in larger batches if needed.
* llama: Add CI to verify all vendored changes have patches (#7066)
Make sure we don't accidentally merge changes in the vendored code
that aren't also reflected in the patches.
* llama: adjust clip patch for mingw utf-16 (#7065)
* llama: adjust clip patch for mingw utf-16
* llama: ensure static linking of runtime libs
Avoid runtime dependencies on non-standard libraries
* runner.go: Enable llamafile (all platforms) and BLAS (Mac OS)
These are two features that are shown on llama.cpp's system info
that are currently different between the two runners. On my test
systems the performance difference is very small to negligible
but it is probably still good to equalize the features.
* llm: Don't add BOS/EOS for tokenize requests
This is consistent with what server.cpp currently does. It affects
things like token processing counts for embedding requests.
* runner.go: Don't cache prompts for embeddings
Our integration with server.cpp implicitly disables prompt caching
because it is not part of the JSON object being parsed, this makes
the Go runner behavior similarly.
Prompt caching has been seen to affect the results of text completions
on certain hardware. The results are not wrong either way but they
are non-deterministic. However, embeddings seem to be affected even
on hardware that does not show this behavior for completions. For
now, it is best to maintain consistency with the existing behavior.
* runner.go: Adjust debug log levels
Add system info printed at startup and quiet down noisier logging.
* llama: fix compiler flag differences (#7082)
Adjust the flags for the new Go server to more closely match the
generate flow
* llama: refine developer docs (#7121)
* llama: doc and example clean up (#7122)
* llama: doc and example clean up
* llama: Move new dockerfile into llama dir
Temporary home until we fully transition to the Go server
* llama: runner doc cleanup
* llama.go: Add description for Tokenize error case
---------
Co-authored-by: Jesse Gross <jesse@ollama.com>
Co-authored-by: Daniel Hiltgen <daniel@ollama.com>
Co-authored-by: Daniel Hiltgen <dhiltgen@users.noreply.github.com>
2024-10-08 15:53:54 +00:00
}