File tree Expand file tree Collapse file tree 1 file changed +3
-3
lines changed Expand file tree Collapse file tree 1 file changed +3
-3
lines changed Original file line number Diff line number Diff line change @@ -73,7 +73,7 @@ set(LLAMA_CUDA_MMQ_Y "64" CACHE STRING "llama: y tile size for mmq CUDA ke
73
73
option (LLAMA_CUDA_FORCE_DMMV "llama: use dmmv instead of mmvq CUDA kernels" OFF )
74
74
set (LLAMA_CUDA_DMMV_X "32" CACHE STRING "llama: x stride for dmmv CUDA kernels" )
75
75
set (LLAMA_CUDA_MMV_Y "1" CACHE STRING "llama: y block size for mmv CUDA kernels" )
76
- option (LLAMA_CUDA_DMMV_F16 "llama: use 16 bit floats for dmmv CUDA kernels " OFF )
76
+ option (LLAMA_CUDA_F16 "llama: use 16 bit floats for some calculations " OFF )
77
77
set (LLAMA_CUDA_KQUANTS_ITER "2" CACHE STRING "llama: iters./thread per block for Q2_K/Q6_K" )
78
78
option (LLAMA_CLBLAST "llama: use CLBlast" OFF )
79
79
option (LLAMA_METAL "llama: use Metal" OFF )
@@ -265,8 +265,8 @@ if (LLAMA_CUBLAS)
265
265
if (DEFINED LLAMA_CUDA_DMMV_Y )
266
266
add_compile_definitions (GGML_CUDA_MMV_Y=${LLAMA_CUDA_DMMV_Y} ) # for backwards compatibility
267
267
endif ()
268
- if (LLAMA_CUDA_DMMV_F16 )
269
- add_compile_definitions (GGML_CUDA_DMMV_F16 )
268
+ if (LLAMA_CUDA_F16 OR LLAMA_CUDA_DMMV_F16 )
269
+ add_compile_definitions (GGML_CUDA_F16 )
270
270
endif ()
271
271
add_compile_definitions (K_QUANTS_PER_ITERATION=${LLAMA_CUDA_KQUANTS_ITER} )
272
272
You can’t perform that action at this time.
0 commit comments