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

PR comments

Browse source
This commit is contained in:
Iwan Kawrakow 2023-06-16 19:36:17 +03:00
parent 31b20758c8
commit 0dc0b6995f
2 changed files with 5 additions and 3 deletions
Download patch file Download diff file Expand all files Collapse all files

6
Makefile
View file

@ -171,11 +171,11 @@ ifdef LLAMA_CUDA_DMMV_Y
else else
NVCCFLAGS += -DGGML_CUDA_DMMV_Y=1 NVCCFLAGS += -DGGML_CUDA_DMMV_Y=1
endif # LLAMA_CUDA_DMMV_Y endif # LLAMA_CUDA_DMMV_Y
#ifdef LLAMA_CUDA_KQUANTS_ITER ifdef LLAMA_CUDA_KQUANTS_ITER
NVCCFLAGS += -DK_QUANTS_PER_ITERATION=$(LLAMA_CUDA_KQUANTS_ITER) NVCCFLAGS += -DK_QUANTS_PER_ITERATION=$(LLAMA_CUDA_KQUANTS_ITER)
#else else
NVCCFLAGS += -DK_QUANTS_PER_ITERATION=2 NVCCFLAGS += -DK_QUANTS_PER_ITERATION=2
#endif endif
ggml-cuda.o: ggml-cuda.cu ggml-cuda.h ggml-cuda.o: ggml-cuda.cu ggml-cuda.h
$(NVCC) $(NVCCFLAGS) $(CXXFLAGS) -Wno-pedantic -c $< -o $@ $(NVCC) $(NVCCFLAGS) $(CXXFLAGS) -Wno-pedantic -c $< -o $@
endif # LLAMA_CUBLAS endif # LLAMA_CUBLAS

2
ggml-cuda.cu
View file

@ -169,6 +169,8 @@ static_assert(sizeof(block_q6_K) == sizeof(ggml_fp16_t) + 13*QK_K/16, "wrong q6_
#ifndef K_QUANTS_PER_ITERATION #ifndef K_QUANTS_PER_ITERATION
#define K_QUANTS_PER_ITERATION 2 #define K_QUANTS_PER_ITERATION 2
#else
static_assert(K_QUANTS_PER_ITERATION == 1 || K_QUANTS_PER_ITERATION == 2, "K_QUANTS_PER_ITERATION must be 1 or 2");
#endif #endif
static __global__ void add_f32(const float * x, const float * y, float * dst, const int k) { static __global__ void add_f32(const float * x, const float * y, float * dst, const int k) {