baalajimaestro/ollama
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity

llm: update llama.cpp commit to 7c26775 (#4896)

Browse source 
* llm: update llama.cpp submodule to `7c26775`

* disable `LLAMA_BLAS` for now

* `-DLLAMA_OPENMP=off`
This commit is contained in:
Jeffrey Morgan 2024-06-17 15:56:16 -04:00 committed by GitHub
parent 4ad0d4d6d3
commit 152fc202f5
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
6 changed files with 31 additions and 29 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

12
llm/generate/gen_darwin.sh
View file

@ -18,7 +18,7 @@ sign() {
fi
}
COMMON_DARWIN_DEFS="-DCMAKE_OSX_DEPLOYMENT_TARGET=11.3 -DLLAMA_METAL_MACOSX_VERSION_MIN=11.3 -DCMAKE_SYSTEM_NAME=Darwin -DLLAMA_METAL_EMBED_LIBRARY=on"
COMMON_DARWIN_DEFS="-DCMAKE_OSX_DEPLOYMENT_TARGET=11.3 -DLLAMA_METAL_MACOSX_VERSION_MIN=11.3 -DCMAKE_SYSTEM_NAME=Darwin -DLLAMA_METAL_EMBED_LIBRARY=on -DLLAMA_OPENMP=off"
case "${GOARCH}" in
"amd64")
@ -27,7 +27,7 @@ case "${GOARCH}" in
# Static build for linking into the Go binary
init_vars
CMAKE_TARGETS="--target llama --target ggml"
CMAKE_DEFS="${COMMON_CPU_DEFS} -DBUILD_SHARED_LIBS=off -DLLAMA_ACCELERATE=off -DLLAMA_AVX=off -DLLAMA_AVX2=off -DLLAMA_AVX512=off -DLLAMA_FMA=off -DLLAMA_F16C=off ${CMAKE_DEFS}"
CMAKE_DEFS="${COMMON_CPU_DEFS} -DBUILD_SHARED_LIBS=off -DLLAMA_BLAS=off -DLLAMA_ACCELERATE=off -DLLAMA_AVX=off -DLLAMA_AVX2=off -DLLAMA_AVX512=off -DLLAMA_FMA=off -DLLAMA_F16C=off ${CMAKE_DEFS}"
BUILD_DIR="../build/darwin/${ARCH}_static"
echo "Building static library"
build
@ -37,7 +37,7 @@ case "${GOARCH}" in
# CPU first for the default library, set up as lowest common denominator for maximum compatibility (including Rosetta)
#
init_vars
CMAKE_DEFS="${COMMON_CPU_DEFS} -DLLAMA_ACCELERATE=off -DLLAMA_AVX=off -DLLAMA_AVX2=off -DLLAMA_AVX512=off -DLLAMA_FMA=off -DLLAMA_F16C=off ${CMAKE_DEFS}"
CMAKE_DEFS="${COMMON_CPU_DEFS} -DLLAMA_ACCELERATE=off -DLLAMA_BLAS=off -DLLAMA_AVX=off -DLLAMA_AVX2=off -DLLAMA_AVX512=off -DLLAMA_FMA=off -DLLAMA_F16C=off ${CMAKE_DEFS}"
BUILD_DIR="../build/darwin/${ARCH}/cpu"
echo "Building LCD CPU"
build
@ -49,7 +49,7 @@ case "${GOARCH}" in
# Approximately 400% faster than LCD on same CPU
#
init_vars
CMAKE_DEFS="${COMMON_CPU_DEFS} -DLLAMA_ACCELERATE=off -DLLAMA_AVX=on -DLLAMA_AVX2=off -DLLAMA_AVX512=off -DLLAMA_FMA=off -DLLAMA_F16C=off ${CMAKE_DEFS}"
CMAKE_DEFS="${COMMON_CPU_DEFS} -DLLAMA_ACCELERATE=off -DLLAMA_BLAS=off -DLLAMA_AVX=on -DLLAMA_AVX2=off -DLLAMA_AVX512=off -DLLAMA_FMA=off -DLLAMA_F16C=off ${CMAKE_DEFS}"
BUILD_DIR="../build/darwin/${ARCH}/cpu_avx"
echo "Building AVX CPU"
build
@ -61,7 +61,7 @@ case "${GOARCH}" in
# Approximately 10% faster than AVX on same CPU
#
init_vars
CMAKE_DEFS="${COMMON_CPU_DEFS} -DLLAMA_ACCELERATE=on -DLLAMA_AVX=on -DLLAMA_AVX2=on -DLLAMA_AVX512=off -DLLAMA_FMA=on -DLLAMA_F16C=on ${CMAKE_DEFS}"
CMAKE_DEFS="${COMMON_CPU_DEFS} -DLLAMA_ACCELERATE=on -DLLAMA_BLAS=off -DLLAMA_AVX=on -DLLAMA_AVX2=on -DLLAMA_AVX512=off -DLLAMA_FMA=on -DLLAMA_F16C=on ${CMAKE_DEFS}"
BUILD_DIR="../build/darwin/${ARCH}/cpu_avx2"
echo "Building AVX2 CPU"
EXTRA_LIBS="${EXTRA_LIBS} -framework Accelerate -framework Foundation"
@ -75,7 +75,7 @@ case "${GOARCH}" in
# Static build for linking into the Go binary
init_vars
CMAKE_TARGETS="--target llama --target ggml"
CMAKE_DEFS="-DCMAKE_OSX_DEPLOYMENT_TARGET=11.3 -DCMAKE_SYSTEM_NAME=Darwin -DBUILD_SHARED_LIBS=off -DCMAKE_SYSTEM_PROCESSOR=${ARCH} -DCMAKE_OSX_ARCHITECTURES=${ARCH} -DLLAMA_METAL=off -DLLAMA_ACCELERATE=off -DLLAMA_AVX=off -DLLAMA_AVX2=off -DLLAMA_AVX512=off -DLLAMA_FMA=off -DLLAMA_F16C=off ${CMAKE_DEFS}"
CMAKE_DEFS="-DCMAKE_OSX_DEPLOYMENT_TARGET=11.3 -DLLAMA_BLAS=off -DCMAKE_SYSTEM_NAME=Darwin -DBUILD_SHARED_LIBS=off -DCMAKE_SYSTEM_PROCESSOR=${ARCH} -DCMAKE_OSX_ARCHITECTURES=${ARCH} -DLLAMA_METAL=off -DLLAMA_ACCELERATE=off -DLLAMA_AVX=off -DLLAMA_AVX2=off -DLLAMA_AVX512=off -DLLAMA_FMA=off -DLLAMA_F16C=off ${CMAKE_DEFS}"
BUILD_DIR="../build/darwin/${ARCH}_static"
echo "Building static library"
build

6
llm/generate/gen_linux.sh
View file

@ -51,7 +51,7 @@ if [ -z "${CUDACXX}" ]; then
export CUDACXX=$(command -v nvcc)
fi
fi
COMMON_CMAKE_DEFS="-DCMAKE_POSITION_INDEPENDENT_CODE=on -DLLAMA_NATIVE=off -DLLAMA_AVX=on -DLLAMA_AVX2=off -DLLAMA_AVX512=off -DLLAMA_FMA=off -DLLAMA_F16C=off"
COMMON_CMAKE_DEFS="-DCMAKE_POSITION_INDEPENDENT_CODE=on -DLLAMA_NATIVE=off -DLLAMA_AVX=on -DLLAMA_AVX2=off -DLLAMA_AVX512=off -DLLAMA_FMA=off -DLLAMA_F16C=off -DLLAMA_OPENMP=off"
source $(dirname $0)/gen_common.sh
init_vars
git_module_setup
@ -64,7 +64,7 @@ if [ -z "${OLLAMA_SKIP_STATIC_GENERATE}" -o "${OLLAMA_CPU_TARGET}" = "static" ];
# Static build for linking into the Go binary
init_vars
CMAKE_TARGETS="--target llama --target ggml"
CMAKE_DEFS="-DBUILD_SHARED_LIBS=off -DLLAMA_NATIVE=off -DLLAMA_AVX=off -DLLAMA_AVX2=off -DLLAMA_AVX512=off -DLLAMA_FMA=off -DLLAMA_F16C=off ${CMAKE_DEFS}"
CMAKE_DEFS="-DBUILD_SHARED_LIBS=off -DLLAMA_NATIVE=off -DLLAMA_AVX=off -DLLAMA_AVX2=off -DLLAMA_AVX512=off -DLLAMA_FMA=off -DLLAMA_F16C=off -DLLAMA_OPENMP=off ${CMAKE_DEFS}"
BUILD_DIR="../build/linux/${ARCH}_static"
echo "Building static library"
build
@ -93,7 +93,7 @@ if [ -z "${OLLAMA_SKIP_CPU_GENERATE}" ]; then
# -DLLAMA_AVX512_VBMI -- 2018 Intel Cannon Lake
# -DLLAMA_AVX512_VNNI -- 2021 Intel Alder Lake
COMMON_CPU_DEFS="-DCMAKE_POSITION_INDEPENDENT_CODE=on -DLLAMA_NATIVE=off"
COMMON_CPU_DEFS="-DCMAKE_POSITION_INDEPENDENT_CODE=on -DLLAMA_NATIVE=off -DLLAMA_OPENMP=off"
if [ -z "${OLLAMA_CPU_TARGET}" -o "${OLLAMA_CPU_TARGET}" = "cpu" ]; then
#
# CPU first for the default library, set up as lowest common denominator for maximum compatibility (including Rosetta)

6
llm/generate/gen_windows.ps1
View file

@ -37,7 +37,8 @@ function init_vars {
}
$script:cmakeDefs = @(
"-DBUILD_SHARED_LIBS=on",
"-DLLAMA_NATIVE=off"
"-DLLAMA_NATIVE=off",
"-DLLAMA_OPENMP=off"
)
$script:commonCpuDefs = @("-DCMAKE_POSITION_INDEPENDENT_CODE=on")
$script:ARCH = $Env:PROCESSOR_ARCHITECTURE.ToLower()
@ -206,7 +207,8 @@ function build_static() {
"-DLLAMA_AVX2=off",
"-DLLAMA_AVX512=off",
"-DLLAMA_F16C=off",
"-DLLAMA_FMA=off")
"-DLLAMA_FMA=off",
"-DLLAMA_OPENMP=off")
$script:buildDir="../build/windows/${script:ARCH}_static"
write-host "Building static library"
build

2
llm/llama.cpp

@ -1 +1 @@
Subproject commit 5921b8f089d3b7bda86aac5a66825df6a6c10603
Subproject commit 7c26775adb579e92b59c82e8084c07a1d0f75e9c

18
llm/patches/01-load-progress.diff
View file

@ -1,8 +1,8 @@
diff --git a/common/common.cpp b/common/common.cpp
index ba1ecf0e..cead57cc 100644
index 73ff0e85..6adb1a92 100644
--- a/common/common.cpp
+++ b/common/common.cpp
@@ -1836,6 +1836,8 @@ struct llama_model_params llama_model_params_from_gpt_params(const gpt_params &
@@ -2447,6 +2447,8 @@ struct llama_model_params llama_model_params_from_gpt_params(const gpt_params &
mparams.use_mmap = params.use_mmap;
mparams.use_mlock = params.use_mlock;
mparams.check_tensors = params.check_tensors;
@ -12,20 +12,20 @@ index ba1ecf0e..cead57cc 100644
mparams.kv_overrides = NULL;
} else {
diff --git a/common/common.h b/common/common.h
index d80344f2..71e84834 100644
index 58ed72f4..0bb2605e 100644
--- a/common/common.h
+++ b/common/common.h
@@ -174,6 +174,13 @@ struct gpt_params {
// multimodal models (see examples/llava)
@@ -180,6 +180,13 @@ struct gpt_params {
std::string mmproj = ""; // path to multimodal projector
std::vector<std::string> image; // path to image file(s)
+
+ // Called with a progress value between 0.0 and 1.0. Pass NULL to disable.
+ // If the provided progress_callback returns true, model loading continues.
+ // If it returns false, model loading is immediately aborted.
+ llama_progress_callback progress_callback = NULL;
+ // context pointer passed to the progress callback
+ void * progress_callback_user_data;
};
void gpt_params_handle_model_default(gpt_params & params);
+
// server params
int32_t port = 8080; // server listens on this network port
int32_t timeout_read = 600; // http read timeout in seconds

16
llm/patches/05-default-pretokenizer.diff
View file

@ -1,8 +1,8 @@
diff --git a/llama.cpp b/llama.cpp
index 40d2ec2c..74f3ee9c 100644
index 61948751..4b72a293 100644
--- a/llama.cpp
+++ b/llama.cpp
@@ -4642,16 +4642,7 @@ static void llm_load_vocab(
@@ -4824,16 +4824,7 @@ static void llm_load_vocab(
// for now, only BPE models have pre-tokenizers
if (vocab.type == LLAMA_VOCAB_TYPE_BPE) {
@ -15,14 +15,14 @@ index 40d2ec2c..74f3ee9c 100644
- LLAMA_LOG_WARN("%s: ************************************ \n", __func__);
- LLAMA_LOG_WARN("%s: \n", __func__);
- vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_DEFAULT;
- } else if (
+ if (
tokenizer_pre == "default") {
- } else if (tokenizer_pre == "default") {
+ if (tokenizer_pre == "default") {
vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_DEFAULT;
} else if (
@@ -4703,7 +4694,8 @@ static void llm_load_vocab(
tokenizer_pre == "smaug-bpe") {
vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_SMAUG;
tokenizer_pre == "llama3" ||
@@ -4888,7 +4879,8 @@ static void llm_load_vocab(
tokenizer_pre == "poro-chat") {
vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_PORO;
} else {
- throw std::runtime_error(format("unknown pre-tokenizer type: '%s'", tokenizer_pre.c_str()));
+ LLAMA_LOG_WARN("%s: missing or unrecognized pre-tokenizer type, using: 'default'\n", __func__);