norohind/ollama
1
0
Fork 0
forked from third-party-mirrors/ollama
Code Pull Requests Activity

fix output

Browse Source
This commit is contained in:
jmorganca 2024-06-09 23:53:40 -04:00
parent 110f37ffb0
commit a632a04426
3 changed files with 25 additions and 15 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
llama/.gitignore vendored
View File

@ -3,5 +3,7 @@
*.lib *.lib
*.exp *.exp
*.dll *.dll
*.so
*.o *.o
ollama_runner_* *.obj
ollama_runner*

34
llama/Makefile
View File

@ -3,21 +3,17 @@ ARCH := $(or $(ARCH), $(shell uname -m))
NVCC := nvcc NVCC := nvcc
HIPCC := "$(HIP_PATH)/bin/hipcc" HIPCC := "$(HIP_PATH)/bin/hipcc"
ifeq ($(UNAME_ARCH),x86_64) ifeq ($(ARCH),x86_64)
ARCH := amd64 ARCH := amd64
endif endif
# Determine object file extension based on OS # Determine object file extension based on OS
ifeq ($(OS),Windows_NT) ifneq (,$(findstring MINGW,$(OS)))
OBJ_EXT := obj OBJ_EXT := obj
SHARED_EXT := dll
else else
OBJ_EXT := o OBJ_EXT := o
endif SHARED_EXT := so
ifeq ($(OS),Windows_NT)
SHARED_EXT := dll
else
SHARED_EXT := so
endif endif
CUDA_SRCS := \ CUDA_SRCS := \
@ -105,15 +101,27 @@ else
all: ollama_runner ollama_runner_avx ollama_runner_avx2 ollama_runner_cuda ollama_runner_rocm all: ollama_runner ollama_runner_avx ollama_runner_avx2 ollama_runner_cuda ollama_runner_rocm
endif endif
%.cuda.$(OBJ_EXT): %.cu %.c %.cpp %.cuda.$(OBJ_EXT): %.cu
nvcc -c $(CUDA_FLAGS) -o $@ $< $(NVCC) -c $(CUDA_FLAGS) -o $@ $<
%.hip.$(OBJ_EXT): %.cu %.c %.cpp %.cuda.$(OBJ_EXT): %.c
$(HIPCC) -c $(HIP_FLAGS) -o $@ $< $(NVCC) -c $(CFLAGS) -o $@ $<
%.cuda.$(OBJ_EXT): %.cpp
$(NVCC) -c $(CXXFLAGS) -o $@ $<
ggml_cuda.$(SHARED_EXT): $(CUDA_OBJS) ggml_cuda.$(SHARED_EXT): $(CUDA_OBJS)
nvcc --shared -lcuda -lcublas -lcudart -lcublasLt $(CUDA_FLAGS) $(CUDA_OBJS) -o $@ nvcc --shared -lcuda -lcublas -lcudart -lcublasLt $(CUDA_FLAGS) $(CUDA_OBJS) -o $@
%.hip.$(OBJ_EXT): %.cu
$(HIPCC) -c $(HIP_FLAGS) -o $@ $<
%.hip.$(OBJ_EXT): %.c
$(HIPCC) -c $(CFLAGS) -o $@ $<
%.hip.$(OBJ_EXT): %.cpp
$(HIPCC) -c $(CXXFLAGS) -o $@ $<
ggml_hipblas.$(SHARED_EXT): $(HIP_OBJS) ggml_hipblas.$(SHARED_EXT): $(HIP_OBJS)
$(HIPCC) --shared -lhipblas -lamdhip64 -lrocblas $(HIP_OBJS) -o $@ $(HIPCC) --shared -lhipblas -lamdhip64 -lrocblas $(HIP_OBJS) -o $@
@ -135,4 +143,4 @@ ollama_runner_rocm: ggml_hipblas.dll
clean: clean:
rm -f $(CUDA_OBJS) $(HIP_OBJS) ggml_cuda.$(SHARED_EXT) ggml_cuda.* ggml_hipblas.* ollama_runner* rm -f $(CUDA_OBJS) $(HIP_OBJS) ggml_cuda.$(SHARED_EXT) ggml_cuda.* ggml_hipblas.* ollama_runner*
.PHONY: all clean ollama_runner_cpu ollama_runner_cpu_avx ollama_runner_cpu_avx2 ollama_runner_cuda ollama_runner_rocm .PHONY: all clean ollama_runner ollama_runner_avx ollama_runner_avx2 ollama_runner_cuda ollama_runner_rocm

2
llama/llama.go
View File

@ -1,6 +1,6 @@
package llama package llama
// #cgo CFLAGS: -std=c11 -DNDEBUG -DLOG_DISABLE_LOGS // #cgo CFLAGS: -std=c11 -DNDEBUG -DLOG_DISABLE_LOGS -Wno-discarded-qualifiers
// #cgo CXXFLAGS: -std=c++11 -DNDEBUG -DLOG_DISABLE_LOGS // #cgo CXXFLAGS: -std=c++11 -DNDEBUG -DLOG_DISABLE_LOGS
// #cgo darwin,arm64 CFLAGS: -DGGML_USE_METAL -DGGML_USE_ACCELERATE -DACCELERATE_NEW_LAPACK -DACCELERATE_LAPACK_ILP64 // #cgo darwin,arm64 CFLAGS: -DGGML_USE_METAL -DGGML_USE_ACCELERATE -DACCELERATE_NEW_LAPACK -DACCELERATE_LAPACK_ILP64
// #cgo darwin,arm64 CXXFLAGS: -DGGML_USE_METAL -DGGML_USE_ACCELERATE -DACCELERATE_NEW_LAPACK -DACCELERATE_LAPACK_ILP64 // #cgo darwin,arm64 CXXFLAGS: -DGGML_USE_METAL -DGGML_USE_ACCELERATE -DACCELERATE_NEW_LAPACK -DACCELERATE_LAPACK_ILP64