matt/ollama37
1
0
Fork 0
mirror of https://github.com/dogkeeper886/ollama37.git synced 2025-12-11 16:26:59 +00:00
Code Issues Packages Projects Releases Wiki Activity

Use build tags to generate accelerated binaries for CUDA and ROCm on Linux.

Browse Source 
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.
This commit is contained in:
65a
2023-10-16 17:41:40 -07:00
committed by Daniel Hiltgen
parent d4cd695759
commit f8ef4439e9
8 changed files with 261 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files

24
llm/llama.cpp/generate_linux_cuda.go Normal file
View File

@@ -0,0 +1,24 @@
//go:build cuda
package llm
//go:generate git submodule init
//go:generate git submodule update --force ggml
//go:generate git -C ggml apply ../patches/0001-add-detokenize-endpoint.patch
//go:generate git -C ggml apply ../patches/0002-34B-model-support.patch
//go:generate git -C ggml apply ../patches/0005-ggml-support-CUDA-s-half-type-for-aarch64-1455-2670.patch
//go:generate git -C ggml apply ../patches/0001-copy-cuda-runtime-libraries.patch
//go:generate rm -rf ggml/build/cuda
//go:generate cmake -S ggml -B ggml/build/cuda -DLLAMA_CUBLAS=on -DLLAMA_ACCELERATE=on -DLLAMA_K_QUANTS=on
//go:generate cmake --build ggml/build/cuda --target server --config Release
//go:generate mv ggml/build/cuda/bin/server ggml/build/cuda/bin/ollama-runner
//go:generate git -C gguf apply ../patches/0001-copy-cuda-runtime-libraries.patch
//go:generate git -C gguf apply ../patches/0001-update-default-log-target.patch
//go:generate rm -rf gguf/build/cuda
//go:generate cmake -S gguf -B gguf/build/cuda -DLLAMA_CUBLAS=on -DLLAMA_ACCELERATE=on -DLLAMA_K_QUANTS=on -DLLAMA_NATIVE=off -DLLAMA_AVX=on -DLLAMA_AVX2=off -DLLAMA_AVX512=off -DLLAMA_FMA=off -DLLAMA_F16C=off
//go:generate cmake --build gguf/build/cuda --target server --config Release
//go:generate mv gguf/build/cuda/bin/server gguf/build/cuda/bin/ollama-runner

25
llm/llama.cpp/generate_linux_rocm.go Normal file
View File

@@ -0,0 +1,25 @@
//go:build rocm
package llm
//go:generate git submodule init
//go:generate git submodule update --force ggml
//go:generate git -C ggml apply ../patches/0001-add-detokenize-endpoint.patch
//go:generate git -C ggml apply ../patches/0002-34B-model-support.patch
//go:generate git -C ggml apply ../patches/0005-ggml-support-CUDA-s-half-type-for-aarch64-1455-2670.patch
//go:generate git -C ggml apply ../patches/0001-copy-cuda-runtime-libraries.patch
//go:generate git submodule update --force gguf
//go:generate git -C gguf apply ../patches/0001-copy-cuda-runtime-libraries.patch
//go:generate git -C gguf apply ../patches/0001-update-default-log-target.patch
//go:generate rm -rf ggml/build/rocm
//go:generate cmake -S ggml -B ggml/build/rocm -DLLAMA_CLBLAST=on -DLLAMA_ACCELERATE=on -DLLAMA_K_QUANTS=on
//go:generate cmake --build ggml/build/rocm --target server --config Release
//go:generate mv ggml/build/rocm/bin/server ggml/build/rocm/bin/ollama-runner
//go:generate rm -rf gguf/build/rocm
//go:generate cmake -S gguf -B gguf/build/rocm -DLLAMA_HIPBLAS=on -DLLAMA_ACCELERATE=on -DLLAMA_K_QUANTS=on -DLLAMA_NATIVE=off -DLLAMA_AVX=on -DLLAMA_AVX2=off -DLLAMA_AVX512=off -DLLAMA_FMA=off -DLLAMA_F16C=off -DCMAKE_C_COMPILER=$ROCM_PATH/llvm/bin/clang -DCMAKE_CXX_COMPILER=$ROCM_PATH/llvm/bin/clang++ -DAMDGPU_TARGETS='gfx803;gfx900;gfx906:xnack-;gfx908:xnack-;gfx90a:xnack+;gfx90a:xnack-;gfx1010;gfx1012;gfx1030;gfx1100;gfx1101;gfx1102' -DGPU_TARGETS='gfx803;gfx900;gfx906:xnack-;gfx908:xnack-;gfx90a:xnack+;gfx90a:xnack-;gfx1010;gfx1012;gfx1030;gfx1100;gfx1101;gfx1102'
//go:generate cmake --build gguf/build/rocm --target server --config Release
//go:generate mv gguf/build/rocm/bin/server gguf/build/rocm/bin/ollama-runner