Commit graph

602 commits

Author SHA1 Message Date
mraiser
4c4c730a0a
Merge branch 'ollama:main' into main 2024-01-27 21:56:11 -05:00
Daniel Hiltgen
e02ecfb6c8
Merge pull request #2116 from dhiltgen/cc_50_80
Add support for CUDA 5.0 cards
2024-01-27 10:28:38 -08:00
Jeffrey Morgan
3ebd6a83fc update submodule to cd4fddb29f81d6a1f6d51a0c016bc6b486d68def 2024-01-25 13:54:11 -08:00
Jeffrey Morgan
a64570dcae
Fix clearing kv cache between requests with the same prompt (#2186)
* Fix clearing kv cache between requests with the same prompt

* fix powershell script
2024-01-25 13:46:20 -08:00
mraiser
a4564232a4
Update gen_linux.sh to find libcudart in separate directory 2024-01-25 09:49:35 -05:00
Michael Yang
cd22855ef8 refactor tensor read 2024-01-24 10:48:31 -08:00
Jeffrey Morgan
4458efb73a
Load all layers on arm64 macOS if model is small enough (#2149) 2024-01-22 17:40:06 -08:00
Daniel Hiltgen
0f5b843319 Refine Accelerate usage on mac
For old macs, accelerate seems to cause crashes, but for
AVX2 capable macs, it does not.
2024-01-22 16:25:56 -08:00
Jeffrey Morgan
ffaf52e1e9 update submodule to 011e8ec577fd135cbc02993d3ea9840c516d6a1c 2024-01-22 15:16:54 -08:00
Daniel Hiltgen
3bc28736cd
Merge pull request #2143 from dhiltgen/llm_verbosity
Refine debug logging for llm
2024-01-22 13:19:16 -08:00
Daniel Hiltgen
730dcfcc7a Refine debug logging for llm
This wires up logging in llama.cpp to always go to stderr, and also
turns up logging if OLLAMA_DEBUG is set.
2024-01-22 12:26:49 -08:00
Daniel Hiltgen
27a2d5af54 Debug logging on init failure 2024-01-22 12:08:22 -08:00
Jeffrey Morgan
5f81a33f43
update submodule to 6f9939d (#2115) 2024-01-22 11:56:40 -08:00
Daniel Hiltgen
5576bb2348
Merge pull request #2130 from dhiltgen/more_faster
Make CPU builds parallel and customizable AMD GPUs
2024-01-21 16:14:12 -08:00
Daniel Hiltgen
ec3764538d Probe GPUs before backend init
Detect potential error scenarios so we can fallback to CPU mode without
hitting asserts.
2024-01-21 15:59:38 -08:00
Daniel Hiltgen
df54c723ae Make CPU builds parallel and customizable AMD GPUs
The linux build now support parallel CPU builds to speed things up.
This also exposes AMD GPU targets as an optional setting for advaced
users who want to alter our default set.
2024-01-21 15:12:21 -08:00
Jeffrey Morgan
89c4aee29e
Unlock mutex when failing to load model (#2117) 2024-01-20 20:54:46 -05:00
Daniel Hiltgen
a447a083f2 Add compute capability 5.0, 7.5, and 8.0 2024-01-20 14:24:05 -08:00
Daniel Hiltgen
681a914990 Add support for CUDA 5.2 cards 2024-01-20 10:48:43 -08:00
Jeffrey Morgan
4c54f0ddeb
sign dylibs on macOS (#2101) 2024-01-19 19:24:11 -05:00
Daniel Hiltgen
6a042438af Switch to local dlopen symbols 2024-01-19 11:37:02 -08:00
Jeffrey Morgan
dc88cc3981
use gzip for runner embedding (#2067) 2024-01-19 13:23:03 -05:00
Daniel Hiltgen
abec7f06e5
Merge pull request #2056 from dhiltgen/slog
Mechanical switch from log to slog
2024-01-18 14:27:24 -08:00
Daniel Hiltgen
fedd705aea Mechanical switch from log to slog
A few obvious levels were adjusted, but generally everything mapped to "info" level.
2024-01-18 14:12:57 -08:00
Daniel Hiltgen
fccdf4c635
Merge pull request #1987 from xyproto/archlinux
Let gpu.go and gen_linux.sh also find CUDA on Arch Linux
2024-01-18 13:32:10 -08:00
Daniel Hiltgen
1b249748ab Add multiple CPU variants for Intel Mac
This also refines the build process for the ext_server build.
2024-01-17 15:08:54 -08:00
Alexander F. Rødseth
cbe2adc78a
Merge branch 'main' into archlinux 2024-01-17 12:50:11 +01:00
Daniel Hiltgen
795674dd90 Bump llama.cpp to b1842 and add new cuda lib dep
Upstream llama.cpp has added a new dependency with the
NVIDIA CUDA Driver Libraries (libcuda.so) which is part of the
driver distribution, not the general cuda libraries, and is not
available as an archive, so we can not statically link it.  This may
introduce some additional compatibility challenges which we'll
need to keep an eye on.
2024-01-16 12:53:52 -08:00
Bruce MacDonald
a897e833b8
do not cache prompt (#2018)
- prompt cache causes inferance to hang after some time
2024-01-16 13:48:05 -05:00
Daniel Hiltgen
8795447dad
Merge pull request #1966 from fpreiss/fpreiss/gen_linux_cuda_detection
improve cuda detection (rel. issue #1704)
2024-01-14 18:00:11 -08:00
Daniel Hiltgen
95ad9a9fc8
Merge pull request #1988 from dhiltgen/fix_intel_mac
Fix typo in arm mac arch script
2024-01-14 08:45:18 -08:00
Daniel Hiltgen
3ca5f69ce8 Fix typo in arm mac arch script 2024-01-14 08:32:57 -08:00
Daniel Hiltgen
cfa6337960
Merge pull request #1982 from dhiltgen/fix_intel_mac
Fix intel mac build
2024-01-14 08:26:46 -08:00
Alexander F. Rødseth
f4bf1d514f Let gpu.go and gen_linux.sh also find CUDA on Arch Linux 2024-01-14 13:40:36 +01:00
Jeffrey Morgan
557110d0ba
Disable mmap with lora layers (#1985) 2024-01-13 23:36:31 -05:00
Daniel Hiltgen
2ecb247276 Fix intel mac build
Make sure we're building an x86 ext_server lib when cross-compiling
2024-01-13 14:46:34 -08:00
Jeffrey Morgan
288ef8ff95
add gcc -lstdc++ flag for linux cpu (#1974) 2024-01-13 03:53:00 -05:00
Jeffrey Morgan
4cf17990f7
use g++ to build libext_server.so on linux (#1972) 2024-01-13 03:12:42 -05:00
Michael Yang
eaed6f8c45 add max context length check 2024-01-12 14:54:07 -08:00
Fabian Preiss
905862e17b improve cuda detection (rel. issue #1704) 2024-01-12 21:59:19 +01:00
Daniel Hiltgen
3773fb6465
Merge pull request #1935 from dhiltgen/cpu_fallback
Fix up the CPU fallback selection
2024-01-11 15:52:32 -08:00
Daniel Hiltgen
7427fa1387 Fix up the CPU fallback selection
The memory changes and multi-variant change had some merge
glitches I missed.  This fixes them so we actually get the cpu llm lib
and best variant for the given system.
2024-01-11 15:27:06 -08:00
Michael Yang
d2be6387c9 fix typo 2024-01-11 14:25:21 -08:00
Michael Yang
d7af35d3d0 import fmt 2024-01-11 14:22:32 -08:00
Michael Yang
defc1dbd6e use x/exp/slices 2024-01-11 14:20:13 -08:00
Daniel Hiltgen
de2fbdec99
Merge pull request #1819 from dhiltgen/multi_variant
Support multiple LLM libs; ROCm v5 and v6; Rosetta, AVX, and AVX2 compatible CPU builds
2024-01-11 14:00:48 -08:00
Michael Yang
f4f939de28
Merge pull request #1552 from jmorganca/mxyng/lint-test
add lint and test on pull_request
2024-01-11 09:37:45 -08:00
Daniel Hiltgen
39928a42e8 Always dynamically load the llm server library
This switches darwin to dynamic loading, and refactors the code now that no
static linking of the library is used on any platform
2024-01-11 08:42:47 -08:00
Daniel Hiltgen
d88c527be3 Build multiple CPU variants and pick the best
This reduces the built-in linux version to not use any vector extensions
which enables the resulting builds to run under Rosetta on MacOS in
Docker.  Then at runtime it checks for the actual CPU vector
extensions and loads the best CPU library available
2024-01-11 08:42:47 -08:00
Jeffrey Morgan
ab6be852c7 revisit memory allocation to account for full kv cache on main gpu 2024-01-11 01:45:31 -05:00
Daniel Hiltgen
8da7bef05f Support multiple variants for a given llm lib type
In some cases we may want multiple variants for a given GPU type or CPU.
This adds logic to have an optional Variant which we can use to select
an optimal library, but also allows us to try multiple variants in case
some fail to load.

This can be useful for scenarios such as ROCm v5 vs v6 incompatibility
or potentially CPU features.
2024-01-10 17:27:51 -08:00
Jeffrey Morgan
b24e8d17b2
Increase minimum CUDA memory allocation overhead and fix minimum overhead for multi-gpu (#1896)
* increase minimum cuda overhead and fix minimum overhead for multi-gpu

* fix multi gpu overhead

* limit overhead to 10% of all gpus

* better wording

* allocate fixed amount before layers

* fixed only includes graph alloc
2024-01-10 19:08:51 -05:00
Jeffrey Morgan
f83881390f revert submodule back to 328b83de23b33240e28f4e74900d1d06726f5eb1 2024-01-10 18:42:39 -05:00
Jeffrey Morgan
224fbf2795 update submodule to commit 1fc2f265ff9377a37fd2c61eae9cd813a3491bea until its main branch is fixed 2024-01-10 17:03:15 -05:00
Jeffrey Morgan
2c6e8f5248
Update submodule to 6efb8eb30e7025b168f3fda3ff83b9b386428ad6 (#1885)
* update submodule to `6efb8eb30e7025b168f3fda3ff83b9b386428ad6`
* unblock condition variable in `update_slots` when closing server
2024-01-10 16:48:38 -05:00
Jeffrey Morgan
34344d801c clean up cmake build directory when cross compiling macOS builds 2024-01-09 17:13:56 -05:00
Jeffrey Morgan
8a8c7e7f8d only build for metal on arm64 2024-01-09 13:51:08 -05:00
Michael Yang
f921e2696e typo 2024-01-09 09:45:42 -08:00
Michael Yang
4a33cede20 remove unused fields and functions 2024-01-09 09:37:40 -08:00
Michael Yang
2bb2bdd5d4 fix lint 2024-01-09 09:36:58 -08:00
Jeffrey Morgan
f387e9631b use runner if cuda alloc won't fit 2024-01-09 00:44:34 -05:00
Jeffrey Morgan
cb534e6ac2 use 10% vram overhead for cuda 2024-01-08 23:17:44 -05:00
Jeffrey Morgan
58ce2d8273 better estimate scratch buffer size 2024-01-08 21:32:44 -05:00
Jeffrey Morgan
18ddf6d57d fix windows build 2024-01-08 20:04:01 -05:00
Jeffrey Morgan
08f1e18965
Offload layers to GPU based on new model size estimates (#1850)
* select layers based on estimated model memory usage

* always account for scratch vram

* dont load +1 layers

* better estmation for graph alloc

* Update gpu/gpu_darwin.go

Co-authored-by: Bruce MacDonald <brucewmacdonald@gmail.com>

* Update llm/llm.go

Co-authored-by: Bruce MacDonald <brucewmacdonald@gmail.com>

* Update llm/llm.go

* add overhead for cuda memory

* Update llm/llm.go

Co-authored-by: Bruce MacDonald <brucewmacdonald@gmail.com>

* fix build error on linux

* address comments

---------

Co-authored-by: Bruce MacDonald <brucewmacdonald@gmail.com>
2024-01-08 16:42:00 -05:00
Jeffrey Morgan
5feec959ad
dont use -Wall in static build (#1833) 2024-01-07 10:39:19 -05:00
Jeffrey Morgan
dbdd50b283
add -DCMAKE_SYSTEM_NAME=Darwin cmake flag (#1832) 2024-01-07 00:46:17 -05:00
Bruce MacDonald
3367b5f3df
remove unused generate patches (#1810) 2024-01-05 11:25:45 -05:00
Daniel Hiltgen
9983fa5f4e Cleaup stale submodule
If the tree has a stale submodule, make sure we clean it up first
2024-01-04 13:40:16 -08:00
Daniel Hiltgen
fac9060da5 Init submodule with new path 2024-01-04 13:00:13 -08:00
Daniel Hiltgen
77d96da94b Code shuffle to clean up the llm dir 2024-01-04 12:12:05 -08:00
Daniel Hiltgen
e9ce91e9a6 Load dynamic cpu lib on windows
On linux, we link the CPU library in to the Go app and fall back to it
when no GPU match is found. On windows we do not link in the CPU library
so that we can better control our dependencies for the CLI.  This fixes
the logic so we correctly fallback to the dynamic CPU library
on windows.
2024-01-04 08:41:41 -08:00
Jeffrey Morgan
c0285158a9 tweak memory requirements error text 2024-01-03 19:47:18 -05:00
Jeffrey Morgan
77a66df72c add macOS memory check for 47B models 2024-01-03 19:46:16 -05:00
Jeffrey Morgan
5b4837f881 remove unused filetype check 2024-01-03 19:45:39 -05:00
Jeffrey Morgan
29340c2e62
update cmake flags for amd64 macOS (#1780)
* update cmake flags for intel macOS

* remove `LLAMA_K_QUANTS`

* put back `CMAKE_OSX_DEPLOYMENT_TARGET` and disable `LLAMA_F16C`
2024-01-03 19:22:15 -05:00
Daniel Hiltgen
d5ec730354
Merge pull request #1779 from dhiltgen/refined_amd_gpu_list
Improve maintainability of Radeon card list
2024-01-03 16:18:57 -08:00
Daniel Hiltgen
ddbfa6fe31 Fix CPU only builds
Go embed doesn't like when there's no matching files, so put
a dummy placeholder in to allow building without any GPU support
If no "server" library is found, it's safely ignored at runtime.
2024-01-03 16:08:34 -08:00
Daniel Hiltgen
16f4603b67 Improve maintainability of Radeon card list
This moves the list of AMD GPUs to an easier to maintain list which
should make it easier to update over time.
2024-01-03 15:16:56 -08:00
Bruce MacDonald
0b3118e0af
fix: relay request opts to loaded llm prediction (#1761) 2024-01-03 12:01:42 -05:00
Daniel Hiltgen
0498f7ce56 Get rid of one-line llama.log
This one log line was triggering a single line llama.log to be generated
in the pwd of the server
2024-01-02 15:36:16 -08:00
Daniel Hiltgen
738a8d12eb Rename the ollama cmakefile 2024-01-02 15:36:16 -08:00
Daniel Hiltgen
d966b730ac Switch windows build to fully dynamic
Refactor where we store build outputs, and support a fully dynamic loading
model on windows so the base executable has no special dependencies thus
doesn't require a special PATH.
2024-01-02 15:36:16 -08:00
Daniel Hiltgen
9a70aecccb Refactor how we augment llama.cpp
This changes the model for llama.cpp inclusion so we're not applying a patch,
but instead have the C++ code directly in the ollama tree, which should make it
easier to refine and update over time.
2024-01-02 15:35:55 -08:00
Jeffrey Morgan
d4ebdadbe7 enable cache_prompt by default 2023-12-27 14:23:42 -05:00
K0IN
10da41d677
Add Cache flag to api (#1642) 2023-12-22 17:16:20 -05:00
Daniel Hiltgen
e5202eb687 Quiet down llama.cpp logging by default
By default builds will now produce non-debug and non-verbose binaries.
To enable verbose logs in llama.cpp and debug symbols in the
native code, set `CGO_CFLAGS=-g`
2023-12-22 08:47:18 -08:00
Daniel Hiltgen
fa24e73b82 Remove CPU build, fixup linux build script 2023-12-21 18:21:31 -08:00
Daniel Hiltgen
325d74985b Fix CPU performance on hyperthreaded systems
The default thread count logic was broken and resulted in 2x the number
of threads as it should on a hyperthreading CPU
resulting in thrashing and poor performance.
2023-12-21 16:23:36 -08:00
Daniel Hiltgen
d9cd3d9667 Revive windows build
The windows native setup still needs some more work, but this gets it building
again and if you set the PATH properly, you can run the resulting exe on a cuda system.
2023-12-20 17:21:54 -08:00
Daniel Hiltgen
7555ea44f8 Revamp the dynamic library shim
This switches the default llama.cpp to be CPU based, and builds the GPU variants
as dynamically loaded libraries which we can select at runtime.

This also bumps the ROCm library to version 6 given 5.7 builds don't work
on the latest ROCm library that just shipped.
2023-12-20 14:45:57 -08:00
Daniel Hiltgen
6558f94ed0 Fix darwin intel build 2023-12-19 13:32:24 -08:00
Daniel Hiltgen
54dbfa4c4a Carry ggml-metal.metal as payload 2023-12-19 09:05:46 -08:00
Daniel Hiltgen
3269535a4c Refine handling of shim presence
This allows the CPU only builds to work on systems with Radeon cards
2023-12-19 09:05:46 -08:00
Daniel Hiltgen
1b991d0ba9 Refine build to support CPU only
If someone checks out the ollama repo and doesn't install the CUDA
library, this will ensure they can build a CPU only version
2023-12-19 09:05:46 -08:00
Daniel Hiltgen
9adca7f711 Bump llama.cpp to b1662 and set n_parallel=1 2023-12-19 09:05:46 -08:00
Daniel Hiltgen
89bbaafa64 Build linux using ubuntu 20.04
This changes the container-based linux build to use an older Ubuntu
distro to improve our compatibility matrix for older user machines
2023-12-19 09:05:46 -08:00
Daniel Hiltgen
35934b2e05 Adapted rocm support to cgo based llama.cpp 2023-12-19 09:05:46 -08:00
65a
f8ef4439e9 Use build tags to generate accelerated binaries for CUDA and ROCm on Linux.
The build tags rocm or cuda must be specified to both go generate and go build.
ROCm builds should have both ROCM_PATH set (and the ROCM SDK present) as well
as CLBlast installed (for GGML) and CLBlast_DIR set in the environment to the
CLBlast cmake directory (likely /usr/lib/cmake/CLBlast). Build tags are also
used to switch VRAM detection between cuda and rocm implementations, using
added "accelerator_foo.go" files which contain architecture specific functions
and variables. accelerator_none is used when no tags are set, and a helper
function addRunner will ignore it if it is the chosen accelerator. Fix go
generate commands, thanks @deadmeu for testing.
2023-12-19 09:05:46 -08:00
Daniel Hiltgen
d4cd695759 Add cgo implementation for llama.cpp
Run the server.cpp directly inside the Go runtime via cgo
while retaining the LLM Go abstractions.
2023-12-19 09:05:46 -08:00
Bruce MacDonald
811b1f03c8 deprecate ggml
- remove ggml runner
- automatically pull gguf models when ggml detected
- tell users to update to gguf in the case automatic pull fails

Co-Authored-By: Jeffrey Morgan <jmorganca@gmail.com>
2023-12-19 09:05:46 -08:00
Jeffrey Morgan
6b5bdfa6c9 update runner submodule 2023-12-18 17:33:46 -05:00
Jeffrey Morgan
c063ee4af0 update runner submodule to fix hipblas build 2023-12-18 15:41:13 -05:00
Jeffrey Morgan
b85982eb91 update runner submodule 2023-12-18 12:43:31 -05:00
Bruce MacDonald
6ee8c80199
restore model load duration on generate response (#1524)
* restore model load duration on generate response

- set model load duration on generate and chat done response
- calculate createAt time when response created

* remove checkpoints predict opts

* Update routes.go
2023-12-14 12:15:50 -05:00
Jeffrey Morgan
31f0551dab
Update runner to support mixtral and mixture of experts (MoE) (#1475) 2023-12-13 17:15:10 -05:00
Michael Yang
4251b342de
Merge pull request #1469 from jmorganca/mxyng/model-types
remove per-model types
2023-12-12 12:27:03 -08:00
Bruce MacDonald
3144e2a439
exponential back-off (#1484) 2023-12-12 12:33:02 -05:00
Bruce MacDonald
c0960e29b5
retry on concurrent request failure (#1483)
- remove parallel
2023-12-12 12:14:35 -05:00
Patrick Devine
910e9401d0
Multimodal support (#1216)
---------

Co-authored-by: Matt Apperson <mattapperson@Matts-MacBook-Pro.local>
2023-12-11 13:56:22 -08:00
Michael Yang
56ffc3023a remove per-model types
mostly replaced by decoding tensors except ggml models which only
support llama
2023-12-11 09:40:21 -08:00
Jeffrey Morgan
fa2f095bd9 fix model name returned by /api/generate being different than the model name provided 2023-12-10 11:42:15 -05:00
Jeffrey Morgan
d9a250e9b5 seek to end of file when decoding older model formats 2023-12-09 21:14:35 -05:00
Jeffrey Morgan
944519ed16 seek to eof for older model binaries 2023-12-09 20:48:57 -05:00
Jeffrey Morgan
2dd040d04c do not use --parallel 2 for old runners 2023-12-09 20:17:33 -05:00
Bruce MacDonald
bbe41ce41a
fix: parallel queueing race condition caused silent failure (#1445)
* fix: queued request failures

- increase parallel requests to 2 to complete queued request, queueing is managed in ollama

* log steam errors
2023-12-09 14:14:02 -05:00
Michael Yang
f1b049fed8
Merge pull request #1377 from jmorganca/mxyng/qwen
update for qwen
2023-12-06 12:31:51 -08:00
Michael Yang
b9495ea162 load projectors 2023-12-05 14:36:12 -08:00
Michael Yang
409bb9674e
Merge pull request #1308 from jmorganca/mxyng/split-from
split from into one or more models
2023-12-05 14:33:03 -08:00
Michael Yang
d3479c07a1
Merge pull request #1250 from jmorganca/mxyng/create-layer
refactor layer creation
2023-12-05 14:32:52 -08:00
Bruce MacDonald
195e3d9dbd
chat api endpoint (#1392) 2023-12-05 14:57:33 -05:00
Jeffrey Morgan
00d06619a1 Revert "chat api (#991)" while context variable is fixed
This reverts commit 7a0899d62d.
2023-12-04 21:16:27 -08:00
Michael Yang
5a5dca13b2 comments 2023-12-04 16:59:23 -08:00
Michael Yang
72e7a49aa9 seek instead of copyn 2023-12-04 16:59:23 -08:00
Michael Yang
2cb0fa7d40 split from into one or more models 2023-12-04 16:59:23 -08:00
Michael Yang
b2816bca67 unnecessary ReadSeeker for DecodeGGML 2023-12-04 16:59:23 -08:00
Bruce MacDonald
7a0899d62d
chat api (#991)
- update chat docs
- add messages chat endpoint
- remove deprecated context and template generate parameters from docs
- context and template are still supported for the time being and will continue to work as expected
- add partial response to chat history
2023-12-04 18:01:06 -05:00
Michael Yang
6deebf2489 update for qwen 2023-12-04 11:38:05 -08:00
Jeffrey Morgan
16a9006306 add back f16c instructions on intel mac 2023-11-26 15:59:49 -05:00
Jeffrey Morgan
9e4a316405 update submodule commit 2023-11-26 14:52:00 -05:00
Jing Zhang
82b9b329ff
windows CUDA support (#1262)
* Support cuda build in Windows
* Enable dynamic NumGPU allocation for Windows
2023-11-24 17:16:36 -05:00
Jongwook Choi
12e8c12d2b
Disable CUDA peer access as a workaround for multi-gpu inference bug (#1261)
When CUDA peer access is enabled, multi-gpu inference will produce
garbage output. This is a known bug of llama.cpp (or nvidia). Until the
upstream bug is fixed, we can disable CUDA peer access temporarily
to ensure correct output.

See #961.
2023-11-24 14:05:57 -05:00
Jeffrey Morgan
d77dde126b consistent cpu instructions on macos and linux 2023-11-22 16:26:46 -05:00
Michael Yang
199941cd15 fix: gguf int type 2023-11-22 11:40:30 -08:00
Michael Yang
a00fac4ec8 update llama.cpp 2023-11-21 09:50:02 -08:00
Jeffrey Morgan
a3fcecf943 only set main_gpu if value > 0 is provided 2023-11-20 19:54:04 -05:00
Michael Yang
19b7a4d715 recent llama.cpp update added kernels for fp32, q5_0, and q5_1 2023-11-20 13:44:31 -08:00
Purinda Gunasekara
be61a81758
main-gpu argument is not getting passed to llamacpp, fixed. (#1192) 2023-11-20 10:52:52 -05:00
Jeffrey Morgan
13ba6df5ab enable cpu instructions on intel macs 2023-11-19 23:20:26 -05:00
Jeffrey Morgan
36a3bbf65f Update llm/llama.go 2023-11-18 21:25:07 -05:00
Bruce MacDonald
43a726149d fix potentially inaccurate error message 2023-11-18 21:25:07 -05:00
Jeffrey Morgan
41434a7cdc build intel mac with correct binary and compile flags 2023-11-16 22:14:51 -05:00
Jeffrey Morgan
5cba29b9d6
JSON mode: add `"format" as an api parameter (#1051)
* add `"format": "json"` as an API parameter
---------
Co-authored-by: Bruce MacDonald <brucewmacdonald@gmail.com>
2023-11-09 16:44:02 -08:00
Bruce MacDonald
1ae84bc2a2
skip gpu if less than 2GB VRAM are available (#1059) 2023-11-09 13:16:16 -08:00
Michael Yang
c5e1bbabda
instead of static number of parameters for each model family, get the real number from the tensors (#1022)
* parse tensor info

* refactor decoder

* return actual parameter count

* explicit rounding

* s/Human/HumanNumber/
2023-11-08 17:55:46 -08:00
Jeffrey Morgan
c44b619428 remove unused fmt.Println 2023-11-03 17:24:58 -07:00
Jeffrey Morgan
17678b7225 Restore system prompt on requests and default num_keep to 0 2023-11-03 13:25:25 -07:00
Jeffrey Morgan
2e53704685
default rope params to 0 for new models (#968) 2023-11-02 08:41:30 -07:00
Michael Yang
642128b75a append LD_LIBRARY_PATH 2023-10-31 15:54:49 -07:00
Jeffrey Morgan
3a1ed9ff70
restore building runner with AVX on by default (#900) 2023-10-27 12:13:44 -07:00
Bruce MacDonald
6d283882b1
catch insufficient permissions nvidia err (#934) 2023-10-27 12:42:40 -04:00
Bruce MacDonald
2665f3c28e
offload 75% of available vram to improve stability (#921) 2023-10-26 20:49:55 -04:00
Jeffrey Morgan
b0c9cd0f3b fix metal assertion errors 2023-10-24 00:32:36 -07:00
Jeffrey Morgan
77f61c6301 update submodule commit 2023-10-24 00:30:27 -07:00
Jeffrey Morgan
f3604534e5 update submodule commit 2023-10-23 23:59:12 -07:00
Michael Yang
0c7a00a264 bump submodules
pin to 9e70cc03229df19ca2d28ce23cc817198f897278 for now since
438c2ca83045a00ef244093d27e9ed41a8cb4ea9 is breaking
2023-10-23 11:17:59 -07:00
Michael Yang
36c160f1c3
Merge pull request #881 from jmorganca/mxyng/ggufv3
ggufv3
2023-10-23 10:50:45 -07:00
Michael Yang
c9167494cb update default log target 2023-10-23 10:44:50 -07:00
Michael Yang
125d0a013a ggufv3
ggufv3 adds support for big endianness, mainly for s390x architecture.
while that's not currently supported for ollama, the change is simple.

loosen version check to be more forward compatible. unless specified,
gguf versions other v1 will be decoded into v2.
2023-10-23 09:35:49 -07:00
Jeffrey Morgan
7ed5a39bc7 simpler check for model loading compatibility errors 2023-10-19 14:50:49 -04:00
Jeffrey Morgan
a7dad24d92
add error for falcon and starcoder vocab compatibility (#844)
add error for falcon and starcoder vocab compatibility
---------
Co-authored-by: Bruce MacDonald <brucewmacdonald@gmail.com>
2023-10-19 12:18:31 -04:00
Michael Yang
235e43d7f6
Merge pull request #833 from discovertomorrow/leadingspace
Fix Issue with Leading Whitespaces in Decoded Context
2023-10-18 13:52:48 -07:00
Arne Müller
730996e530 use TrimPrefix instead of TrimLeft 2023-10-18 22:51:30 +02:00
Arne Müller
ce6197a8e0 removed redundant strings.CutPrefix from Decode 2023-10-18 22:47:20 +02:00
Arne Müller
46b9953f32 use strings.TrimLeft to remove spaces 2023-10-18 22:41:19 +02:00
Bruce MacDonald
565648f3f7
relay CUDA errors to the client (#825) 2023-10-18 15:36:56 -04:00
Arne Müller
90c49bed57 moved removal of leading space into Predict 2023-10-18 20:08:26 +02:00
Arne Müller
5dc0cff459 fix whitespace removal 2023-10-18 08:15:27 +02:00
Michael Yang
08b0e04f40
Merge pull request #813 from jmorganca/mxyng/llama
refactor llm/llama.go
2023-10-17 14:05:58 -07:00
Michael Yang
b36b0b71f8 use cut prefix 2023-10-17 14:01:39 -07:00
Michael Yang
094df37563 remove unused struct 2023-10-17 14:01:38 -07:00
Bruce MacDonald
f3648fd206
Update llama.cpp gguf to latest (#710) 2023-10-17 16:55:16 -04:00
Bruce MacDonald
bd93a94abd
fix MB VRAM log output (#824) 2023-10-17 15:35:16 -04:00
Michael Yang
f55bdb6f10
Merge pull request #799 from deichbewohner/jsonmarshaling
Fix JSON Marshal Escaping for Special Characters
2023-10-17 08:46:02 -07:00
Michael Yang
2870a9bfc8
Merge pull request #812 from jmorganca/mxyng/fix-format-string
fix: wrong format string type
2023-10-17 08:40:49 -07:00
Arne Müller
8fa3f366ad Removed newline trimming and used buffer directly in POST request. 2023-10-17 08:17:35 +02:00
Michael Yang
fddb303f23 fix: format string wrong type 2023-10-16 16:14:28 -07:00
Michael Yang
cb4a80b693 fix: regression unsupported metal types
omitting `--n-gpu-layers` means use metal on macos which isn't correct
since ollama uses `num_gpu=0` to explicitly disable gpu for file types
that are not implemented in metal
2023-10-16 14:37:20 -07:00
Arne Müller
ee94693b1a handling unescaped json marshaling 2023-10-16 11:15:55 +02:00
Michael Yang
11d82d7b9b update checkvram 2023-10-13 14:47:29 -07:00
Michael Yang
36fe2deebf only check system memory on macos 2023-10-13 14:47:29 -07:00
Michael Yang
4a8931f634 check total (system + video) memory 2023-10-13 14:47:29 -07:00
Michael Yang
bd6e38fb1a refactor memory check 2023-10-13 14:47:29 -07:00
Michael Yang
92189a5855 fix memory check 2023-10-13 14:47:29 -07:00
Michael Yang
d790bf9916
Merge pull request #783 from jmorganca/mxyng/fix-gpu-offloading
fix: offloading on low end GPUs
2023-10-13 14:36:44 -07:00
Michael Yang
35afac099a do not use gpu binary when num_gpu == 0 2023-10-13 14:32:12 -07:00
Michael Yang
811c3d1900 no gpu if vram < 2GB 2023-10-13 14:32:12 -07:00
Bruce MacDonald
6fe178134d
improve api error handling (#781)
- remove new lines from llama.cpp error messages relayed to client
- check api option types and return error on wrong type
- change num layers from 95% VRAM to 92% VRAM
2023-10-13 16:57:10 -04:00
Bruce MacDonald
56497663c8
relay model runner error message to client (#720)
* give direction to user when runner fails
* also relay errors from timeout
* increase timeout to 3 minutes
2023-10-12 11:16:37 -04:00
Michael Yang
b599946b74 add format bytes 2023-10-11 14:08:23 -07:00
Bruce MacDonald
77295f716e
prevent waiting on exited command (#752)
* prevent waiting on exited command
* close llama runner once
2023-10-11 12:32:13 -04:00
Bruce MacDonald
f2ba1311aa
improve vram safety with 5% vram memory buffer (#724)
* check free memory not total
* wait for subprocess to exit
2023-10-10 16:16:09 -04:00
Jeffrey Morgan
ab0668293c llm: fix build on amd64 2023-10-06 14:39:54 -07:00
Bruce MacDonald
5d22319a2c
rename server subprocess (#700)
- this makes it easier to see that the subprocess is associated with ollama
2023-10-06 10:15:42 -04:00
Bruce MacDonald
d06bc0cb6e
enable q8, q5, 5_1, and f32 for linux gpu (#699) 2023-10-05 12:53:47 -04:00
Bruce MacDonald
9e2de1bd2c
increase streaming buffer size (#692) 2023-10-04 14:09:00 -04:00
Michael Yang
c02c0cd483 starcoder 2023-10-02 19:56:51 -07:00
Bruce MacDonald
b1f7123301
clean up num_gpu calculation code (#673) 2023-10-02 14:53:42 -04:00
Bruce MacDonald
1fbf3585d6
Relay default values to llama runner (#672)
* include seed in params for llama.cpp server and remove empty filter for temp

* relay default predict options to llama.cpp

- reorganize options to match predict request for readability

* omit empty stop

---------

Co-authored-by: hallh <hallh@users.noreply.github.com>
2023-10-02 14:53:16 -04:00
Bruce MacDonald
9771b1ec51
windows runner fixes (#637) 2023-09-29 11:47:55 -04:00
Michael Yang
f40b3de758 use int64 consistently 2023-09-28 11:07:24 -07:00
Bruce MacDonald
86279f4ae3
unbound max num gpu layers (#591)
---------

Co-authored-by: Michael Yang <mxyng@pm.me>
2023-09-25 18:36:46 -04:00
Michael Yang
058d0cd04b silence warm up log 2023-09-21 14:53:33 -07:00
Michael Yang
ee1c994d15
update submodule (#567) 2023-09-21 16:22:23 -04:00
Bruce MacDonald
4cba75efc5
remove tmp directories created by previous servers (#559)
* remove tmp directories created by previous servers

* clean up on server stop

* Update routes.go

* Update server/routes.go

Co-authored-by: Jeffrey Morgan <jmorganca@gmail.com>

* create top-level temp ollama dir

* check file exists before creating

---------

Co-authored-by: Jeffrey Morgan <jmorganca@gmail.com>
Co-authored-by: Michael Yang <mxyng@pm.me>
2023-09-21 20:38:49 +01:00
Michael Yang
a9ed7cc6aa rename generate.go 2023-09-20 14:42:17 -07:00
Michael Yang
6c6a31a1e8 embed libraries using cmake 2023-09-20 14:41:57 -07:00
Bruce MacDonald
fc6ec356fc remove libcuda.so 2023-09-20 20:36:14 +01:00
Bruce MacDonald
1255bc9b45 only package 11.8 runner 2023-09-20 20:00:41 +01:00
Bruce MacDonald
b9bb5ca288 use cuda_version 2023-09-20 17:58:16 +01:00
Bruce MacDonald
4e8be787c7 pack in cuda libs 2023-09-20 17:40:42 +01:00
Bruce MacDonald
66003e1d05
subprocess improvements (#524)
* subprocess improvements

- increase start-up timeout
- when runner fails to start fail rather than timing out
- try runners in order rather than choosing 1 runner
- embed metal runner in metal dir rather than gpu
- refactor logging and error messages

* Update llama.go

* Update llama.go

* simplify by using glob
2023-09-18 15:16:32 -04:00
Bruce MacDonald
2540c9181c
support for packaging in multiple cuda runners (#509)
* enable packaging multiple cuda versions
* use nvcc cuda version if available

---------

Co-authored-by: Michael Yang <mxyng@pm.me>
2023-09-14 15:08:13 -04:00
Michael Yang
d028853879 fix: add falcon.go 2023-09-13 14:47:37 -07:00
Michael Yang
949553db23
Merge pull request #519 from jmorganca/mxyng/decode
Mxyng/decode
2023-09-13 12:43:57 -07:00
Michael Yang
0c5a454361 fix model type for 70b 2023-09-12 15:12:59 -07:00
Bruce MacDonald
f59c4d03f7
fix ggml arm64 cuda build (#520) 2023-09-12 17:06:48 -04:00
Michael Yang
7dee25a07f fix falcon decode
get model and file type from bin file
2023-09-12 12:34:53 -07:00
Bruce MacDonald
f221637053
first pass at linux gpu support (#454)
* linux gpu support
* handle multiple gpus
* add cuda docker image (#488)
---------

Co-authored-by: Michael Yang <mxyng@pm.me>
2023-09-12 11:04:35 -04:00
Bruce MacDonald
09dd2aeff9
GGUF support (#441) 2023-09-07 13:55:37 -04:00
Jeffrey Morgan
61dda6a5e0 set minimum CMAKE_OSX_DEPLOYMENT_TARGET to 11.0 2023-09-06 19:56:50 -04:00
Jeffrey Morgan
7de300856b use osPath in gpu check 2023-09-05 21:52:21 -04:00
Jeffrey Morgan
213ffdb548 macos amd64 compatibility fixes 2023-09-05 21:33:31 -04:00
Bruce MacDonald
d18282bfda
metal: add missing barriers for mul-mat (#469) 2023-09-05 19:37:13 -04:00
Michael Yang
2bc06565c7 fix empty response 2023-09-05 15:03:24 -07:00
Michael Yang
7b5aefb427
Merge pull request #462 from jmorganca/mxyng/rm-marshal-prompt
remove marshalPrompt which is no longer needed
2023-09-05 11:48:41 -07:00
Jeffrey Morgan
7fa6e51686
generate binary dependencies based on GOARCH on macos (#459) 2023-09-05 12:53:57 -04:00
Michael Yang
59a705525c fix not forwarding last token 2023-09-03 17:46:50 -04:00
Michael Yang
5d3f314b0b remove marshalPrompt which is no longer needed 2023-09-03 17:01:05 -04:00
Bruce MacDonald
f964aea9a2 remove test not applicate to subprocess 2023-08-30 16:36:11 -04:00
Bruce MacDonald
42998d797d
subprocess llama.cpp server (#401)
* remove c code
* pack llama.cpp
* use request context for llama_cpp
* let llama_cpp decide the number of threads to use
* stop llama runner when app stops
* remove sample count and duration metrics
* use go generate to get libraries
* tmp dir for running llm
2023-08-30 16:35:03 -04:00
Quinn Slack
f4432e1dba
treat stop as stop sequences, not exact tokens (#442)
The `stop` option to the generate API is a list of sequences that should cause generation to stop. Although these are commonly called "stop tokens", they do not necessarily correspond to LLM tokens (per the LLM's tokenizer). For example, if the caller sends a generate request with `"stop":["\n"]`, then generation should stop on any token containing `\n` (and trim `\n` from the output), not just if the token exactly matches `\n`. If `stop` were interpreted strictly as LLM tokens, then it would require callers of the generate API to know the LLM's tokenizer and enumerate many tokens in the `stop` list.

Fixes https://github.com/jmorganca/ollama/issues/295.
2023-08-30 11:53:42 -04:00
Michael Yang
7df342a6ea
Merge pull request #421 from jmorganca/mxyng/f16-metal
allow F16 to use metal
2023-08-29 06:32:59 -07:00
Michael Yang
e82fcf30c6
Merge pull request #420 from jmorganca/mxyng/34b-mem-check
add 34b to mem check
2023-08-26 14:15:52 -07:00
Michael Yang
b25dd1795d allow F16 to use metal
warning F16 uses significantly more memory than quantized model so the
standard requires don't apply.
2023-08-26 08:38:48 -07:00
Michael Yang
304f2b6c96 add 34b to mem check 2023-08-26 08:29:21 -07:00
Jeffrey Morgan
177b69a211 add missing entries for 34B 2023-08-25 18:35:35 -07:00
Michael Yang
7a378f8b66 patch llama.cpp for 34B 2023-08-25 10:06:55 -07:00
Michael Yang
b1cececb8e add 34b model type 2023-08-24 10:35:44 -07:00
Michael Yang
5ca05c2e88 fix ModelType() 2023-08-18 11:23:38 -07:00
Michael Yang
a894cc792d model and file type as strings 2023-08-17 12:08:04 -07:00
Michael Yang
4dcf5c3e0b
Merge pull request #349 from jmorganca/close-files
close open files
2023-08-14 16:15:58 -07:00
Michael Yang
e26085b921 close open files 2023-08-14 16:08:06 -07:00
Michael Yang
f7b613332c update llama.cpp 2023-08-14 15:47:00 -07:00
Bruce MacDonald
4b2d366c37 Update llama.go 2023-08-14 12:55:50 -03:00
Bruce MacDonald
56fd4e4ef2 log embedding eval timing 2023-08-14 12:51:31 -03:00
Jeffrey Morgan
22885aeaee update llama.cpp to f64d44a 2023-08-12 22:47:15 -04:00
Michael Yang
6ed991c8e2 ggml: fix off by one error
remove used Unknown FileType
2023-08-11 10:45:22 -07:00
Michael Yang
6de5d032e1 implement loading ggml lora adapters through the modelfile 2023-08-10 09:23:39 -07:00
Michael Yang
d791df75dd check memory requirements before loading 2023-08-10 09:23:11 -07:00
Michael Yang
020a3b3530 disable gpu for q5_0, q5_1, q8_0 quants 2023-08-10 09:23:11 -07:00
Michael Yang
fccf8d179f partial decode ggml bin for more info 2023-08-10 09:23:10 -07:00