vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Merge branch 'ggerganov:master' into master

Browse source
This commit is contained in:
Yazan Agha-Schrader 2023-11-28 09:43:39 +01:00 • committed by GitHub
commit c96c458fe5
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
6 changed files with 31 additions and 35 deletions
Download patch file Download diff file Expand all files Collapse all files

4
CMakeLists.txt
View file

@ -43,6 +43,7 @@ else()
endif() endif()
# general # general
option(BUILD_SHARED_LIBS "build shared libraries" OFF)
option(LLAMA_STATIC "llama: static link libraries" OFF) option(LLAMA_STATIC "llama: static link libraries" OFF)
option(LLAMA_NATIVE "llama: enable -march=native flag" ON) option(LLAMA_NATIVE "llama: enable -march=native flag" ON)
option(LLAMA_LTO "llama: enable link time optimization" OFF) option(LLAMA_LTO "llama: enable link time optimization" OFF)
@ -100,6 +101,9 @@ option(LLAMA_BUILD_TESTS "llama: build tests" ${LLAMA_STANDALO
option(LLAMA_BUILD_EXAMPLES "llama: build examples" ${LLAMA_STANDALONE}) option(LLAMA_BUILD_EXAMPLES "llama: build examples" ${LLAMA_STANDALONE})
option(LLAMA_BUILD_SERVER "llama: build server example" ON) option(LLAMA_BUILD_SERVER "llama: build server example" ON)
# Required for relocatable CMake package
include(${CMAKE_CURRENT_SOURCE_DIR}/scripts/build-info.cmake)
# #
# Compile flags # Compile flags
# #

2
common/CMakeLists.txt
View file

@ -26,7 +26,7 @@ add_custom_command(
COMMENT "Generating build details from Git" COMMENT "Generating build details from Git"
COMMAND ${CMAKE_COMMAND} -DMSVC=${MSVC} -DCMAKE_C_COMPILER_VERSION=${CMAKE_C_COMPILER_VERSION} COMMAND ${CMAKE_COMMAND} -DMSVC=${MSVC} -DCMAKE_C_COMPILER_VERSION=${CMAKE_C_COMPILER_VERSION}
-DCMAKE_C_COMPILER_ID=${CMAKE_C_COMPILER_ID} -DCMAKE_VS_PLATFORM_NAME=${CMAKE_VS_PLATFORM_NAME} -DCMAKE_C_COMPILER_ID=${CMAKE_C_COMPILER_ID} -DCMAKE_VS_PLATFORM_NAME=${CMAKE_VS_PLATFORM_NAME}
-DCMAKE_C_COMPILER=${CMAKE_C_COMPILER} -P "${CMAKE_CURRENT_SOURCE_DIR}/../scripts/build-info.cmake" -DCMAKE_C_COMPILER=${CMAKE_C_COMPILER} -P "${CMAKE_CURRENT_SOURCE_DIR}/../scripts/gen-build-info-cpp.cmake"
WORKING_DIRECTORY "${CMAKE_CURRENT_SOURCE_DIR}/.." WORKING_DIRECTORY "${CMAKE_CURRENT_SOURCE_DIR}/.."
DEPENDS "${CMAKE_CURRENT_SOURCE_DIR}/build-info.cpp.in" ${GIT_INDEX} DEPENDS "${CMAKE_CURRENT_SOURCE_DIR}/build-info.cpp.in" ${GIT_INDEX}
VERBATIM VERBATIM

2
ggml.c
View file

@ -9373,7 +9373,7 @@ static bool ggml_compute_forward_mul_mat_use_blas(
// TODO: find the optimal values for these // TODO: find the optimal values for these
if (ggml_is_contiguous(src0) && if (ggml_is_contiguous(src0) &&
ggml_is_contiguous(src1) && ggml_is_contiguous(src1) &&
src0->type == GGML_TYPE_F32 && //src0->type == GGML_TYPE_F32 &&
src1->type == GGML_TYPE_F32 && src1->type == GGML_TYPE_F32 &&
(ne0 >= 32 && ne1 >= 32 && ne10 >= 32)) { (ne0 >= 32 && ne1 >= 32 && ne10 >= 32)) {

12
llama.cpp
View file

@ -5550,18 +5550,8 @@ static int llama_decode_internal(
n_threads = std::min(4, n_threads); n_threads = std::min(4, n_threads);
} }
// If all tensors can be run on the GPU then using more than 1 thread is detrimental.
const bool full_offload_supported =
model.arch == LLM_ARCH_LLAMA ||
model.arch == LLM_ARCH_BAICHUAN ||
model.arch == LLM_ARCH_FALCON ||
model.arch == LLM_ARCH_REFACT ||
model.arch == LLM_ARCH_MPT ||
model.arch == LLM_ARCH_STARCODER ||
model.arch == LLM_ARCH_STABLELM;
const bool fully_offloaded = model.n_gpu_layers >= (int) hparams.n_layer + 3; const bool fully_offloaded = model.n_gpu_layers >= (int) hparams.n_layer + 3;
if (ggml_cpu_has_cublas() && full_offload_supported && fully_offloaded) { if (ggml_cpu_has_cublas() && fully_offloaded) {
n_threads = 1; n_threads = 1;
} }

22
scripts/build-info.cmake
View file

@ -1,5 +1,3 @@
set(TEMPLATE_FILE "${CMAKE_CURRENT_SOURCE_DIR}/common/build-info.cpp.in")
set(OUTPUT_FILE "${CMAKE_CURRENT_SOURCE_DIR}/common/build-info.cpp")
set(BUILD_NUMBER 0) set(BUILD_NUMBER 0)
set(BUILD_COMMIT "unknown") set(BUILD_COMMIT "unknown")
set(BUILD_COMPILER "unknown") set(BUILD_COMPILER "unknown")
@ -58,23 +56,3 @@ else()
) )
set(BUILD_TARGET ${OUT}) set(BUILD_TARGET ${OUT})
endif() endif()
# Only write the build info if it changed
if(EXISTS ${OUTPUT_FILE})
file(READ ${OUTPUT_FILE} CONTENTS)
string(REGEX MATCH "LLAMA_COMMIT = \"([^\"]*)\";" _ ${CONTENTS})
set(OLD_COMMIT ${CMAKE_MATCH_1})
string(REGEX MATCH "LLAMA_COMPILER = \"([^\"]*)\";" _ ${CONTENTS})
set(OLD_COMPILER ${CMAKE_MATCH_1})
string(REGEX MATCH "LLAMA_BUILD_TARGET = \"([^\"]*)\";" _ ${CONTENTS})
set(OLD_TARGET ${CMAKE_MATCH_1})
if (
NOT OLD_COMMIT STREQUAL BUILD_COMMIT OR
NOT OLD_COMPILER STREQUAL BUILD_COMPILER OR
NOT OLD_TARGET STREQUAL BUILD_TARGET
)
configure_file(${TEMPLATE_FILE} ${OUTPUT_FILE})
endif()
else()
configure_file(${TEMPLATE_FILE} ${OUTPUT_FILE})
endif()

24
scripts/gen-build-info-cpp.cmake Normal file
View file

@ -0,0 +1,24 @@
include(${CMAKE_CURRENT_SOURCE_DIR}/scripts/build-info.cmake)
set(TEMPLATE_FILE "${CMAKE_CURRENT_SOURCE_DIR}/common/build-info.cpp.in")
set(OUTPUT_FILE "${CMAKE_CURRENT_SOURCE_DIR}/common/build-info.cpp")
# Only write the build info if it changed
if(EXISTS ${OUTPUT_FILE})
file(READ ${OUTPUT_FILE} CONTENTS)
string(REGEX MATCH "LLAMA_COMMIT = \"([^\"]*)\";" _ ${CONTENTS})
set(OLD_COMMIT ${CMAKE_MATCH_1})
string(REGEX MATCH "LLAMA_COMPILER = \"([^\"]*)\";" _ ${CONTENTS})
set(OLD_COMPILER ${CMAKE_MATCH_1})
string(REGEX MATCH "LLAMA_BUILD_TARGET = \"([^\"]*)\";" _ ${CONTENTS})
set(OLD_TARGET ${CMAKE_MATCH_1})
if (
NOT OLD_COMMIT STREQUAL BUILD_COMMIT OR
NOT OLD_COMPILER STREQUAL BUILD_COMPILER OR
NOT OLD_TARGET STREQUAL BUILD_TARGET
)
configure_file(${TEMPLATE_FILE} ${OUTPUT_FILE})
endif()
else()
configure_file(${TEMPLATE_FILE} ${OUTPUT_FILE})
endif()