From fc6ec356fc018704ff49823a624e95069544408f Mon Sep 17 00:00:00 2001 From: Bruce MacDonald Date: Wed, 20 Sep 2023 20:36:14 +0100 Subject: [PATCH] remove libcuda.so --- docs/development.md | 2 +- llm/llama.cpp/generate_linux.go | 3 +-- 2 files changed, 2 insertions(+), 3 deletions(-) diff --git a/docs/development.md b/docs/development.md index 98a81b4b..e1b00bac 100644 --- a/docs/development.md +++ b/docs/development.md @@ -35,5 +35,5 @@ Now you can run `ollama`: ## Building on Linux with GPU support - Install cmake and nvidia-cuda-toolkit -- run `CUDA_PATH=/path/to/libcuda.so CUBLAS_PATH=/path/to/libcublas.so CUDART_PATH=/path/to/libcudart.so CUBLASLT_PATH=/path/to/libcublasLt.so go generate ./...` +- run `CUBLAS_PATH=/path/to/libcublas.so CUDART_PATH=/path/to/libcudart.so CUBLASLT_PATH=/path/to/libcublasLt.so go generate ./...` - run `go build .` diff --git a/llm/llama.cpp/generate_linux.go b/llm/llama.cpp/generate_linux.go index c9303c4f..a3ee703d 100644 --- a/llm/llama.cpp/generate_linux.go +++ b/llm/llama.cpp/generate_linux.go @@ -19,8 +19,7 @@ package llm //go:generate cmake --build ggml/build/cuda --target server --config Release //go:generate cmake -S gguf -B gguf/build/cuda -DLLAMA_CUBLAS=on -DLLAMA_ACCELERATE=on -DLLAMA_K_QUANTS=on //go:generate cmake --build gguf/build/cuda --target server --config Release -//go:generate cp --dereference ${CUDA_PATH} ggml/build/cuda/bin/libcuda.so -//go:generate cp --dereference ${CUDA_PATH} gguf/build/cuda/bin/libcuda.so + //go:generate cp --dereference ${CUBLAS_PATH} ggml/build/cuda/bin/libcublas.so.11 //go:generate cp --dereference ${CUBLAS_PATH} gguf/build/cuda/bin/libcublas.so.11 //go:generate cp --dereference ${CUDART_PATH} ggml/build/cuda/bin/libcudart.so.11.0