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

ggml : remove K_QUANTS_PER_ITERATION (#8306)

Browse source 
ggml-ci
This commit is contained in:
Georgi Gerganov 2024-07-10 15:23:12 +03:00 • committed by GitHub
parent 91deef4606
commit 117f7adbd9
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
13 changed files with 110 additions and 282 deletions
Download patch file Download diff file Expand all files Collapse all files

2
ggml/CMakeLists.txt
View file

@ -113,8 +113,6 @@ option(GGML_CUDA_FORCE_CUBLAS "ggml: always use cuBLAS instead of
set (GGML_CUDA_DMMV_X "32" CACHE STRING "ggml: x stride for dmmv CUDA kernels")
set (GGML_CUDA_MMV_Y "1" CACHE STRING "ggml: y block size for mmv CUDA kernels")
option(GGML_CUDA_F16 "ggml: use 16 bit floats for some calculations" OFF)
set (GGML_CUDA_KQUANTS_ITER "2" CACHE STRING
"ggml: iters./thread per block for Q2_K/Q6_K")
set (GGML_CUDA_PEER_MAX_BATCH_SIZE "128" CACHE STRING
"ggml: max. batch size for using peer access")
option(GGML_CUDA_NO_PEER_COPY "ggml: do not use peer to peer copies" OFF)