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

Fix FP32/FP16 build instructions

Browse source
This commit is contained in:
Anas Ahouzi 2024-04-19 05:52:58 -07:00
parent 637e9a86c2
commit e1f6992d7c
1 changed files with 13 additions and 8 deletions
Download patch file Download diff file Expand all files Collapse all files

21
README-sycl.md
View file

@ -229,12 +229,12 @@ source /opt/intel/oneapi/setvars.sh
# Build LLAMA with MKL BLAS acceleration for intel GPU # Build LLAMA with MKL BLAS acceleration for intel GPU
mkdir -p build && cd build mkdir -p build && cd build
# Option 1: Use FP16 for better performance in long-prompt inference # Option 1: Use FP32 by default
#cmake .. -DLLAMA_SYCL=ON -DCMAKE_C_COMPILER=icx -DCMAKE_CXX_COMPILER=icpx -DLLAMA_SYCL_F16=ON
# Option 2: Use FP32 by default
cmake .. -DLLAMA_SYCL=ON -DCMAKE_C_COMPILER=icx -DCMAKE_CXX_COMPILER=icpx cmake .. -DLLAMA_SYCL=ON -DCMAKE_C_COMPILER=icx -DCMAKE_CXX_COMPILER=icpx
# Option 2: Use FP16 for better performance in long-prompt inference
cmake .. -DLLAMA_SYCL=ON -DCMAKE_C_COMPILER=icx -DCMAKE_CXX_COMPILER=icpx -DLLAMA_SYCL_F16=ON
#build all binary #build all binary
cmake --build . --config Release -j -v cmake --build . --config Release -j -v
``` ```
@ -250,12 +250,12 @@ export CPLUS_INCLUDE_DIR=/path/to/oneMKL/include:$CPLUS_INCLUDE_DIR
# Build LLAMA with Nvidia BLAS acceleration through SYCL # Build LLAMA with Nvidia BLAS acceleration through SYCL
mkdir -p build && cd build mkdir -p build && cd build
# Option 1: Use FP16 for better performance in long-prompt inference # Option 1: Use FP32 by default
cmake .. -DLLAMA_SYCL=ON -DLLAMA_SYCL_TARGET=NVIDIA -DCMAKE_C_COMPILER=icx -DCMAKE_CXX_COMPILER=icpx -DLLAMA_SYCL_F16=ON
# Option 2: Use FP32 by default
cmake .. -DLLAMA_SYCL=ON -DLLAMA_SYCL_TARGET=NVIDIA -DCMAKE_C_COMPILER=icx -DCMAKE_CXX_COMPILER=icpx cmake .. -DLLAMA_SYCL=ON -DLLAMA_SYCL_TARGET=NVIDIA -DCMAKE_C_COMPILER=icx -DCMAKE_CXX_COMPILER=icpx
# Option 2: Use FP16 for better performance in long-prompt inference
cmake .. -DLLAMA_SYCL=ON -DLLAMA_SYCL_TARGET=NVIDIA -DCMAKE_C_COMPILER=icx -DCMAKE_CXX_COMPILER=icpx -DLLAMA_SYCL_F16=ON
#build all binary #build all binary
cmake --build . --config Release -j -v cmake --build . --config Release -j -v
@ -416,8 +416,13 @@ mkdir -p build
cd build cd build
@call "C:\Program Files (x86)\Intel\oneAPI\setvars.bat" intel64 --force @call "C:\Program Files (x86)\Intel\oneAPI\setvars.bat" intel64 --force
# Option 1: Use FP32 by default
cmake -G "MinGW Makefiles" .. -DLLAMA_SYCL=ON -DCMAKE_C_COMPILER=icx -DCMAKE_CXX_COMPILER=icx -DCMAKE_BUILD_TYPE=Release
# Option 2: Or FP16
cmake -G "MinGW Makefiles" .. -DLLAMA_SYCL=ON -DCMAKE_C_COMPILER=icx -DCMAKE_CXX_COMPILER=icx -DCMAKE_BUILD_TYPE=Release -DLLAMA_SYCL_F16=ON cmake -G "MinGW Makefiles" .. -DLLAMA_SYCL=ON -DCMAKE_C_COMPILER=icx -DCMAKE_CXX_COMPILER=icx -DCMAKE_BUILD_TYPE=Release -DLLAMA_SYCL_F16=ON
make -j make -j
``` ```