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

CUDA: quantized KV support for FA vec

Browse source
This commit is contained in:
Johannes Gäßler 2024-05-21 19:38:25 +02:00
parent 10b1e45876
commit 672244a88b
11 changed files with 826 additions and 142 deletions
Download patch file Download diff file Expand all files Collapse all files

4
CMakeLists.txt
View file

@ -106,6 +106,7 @@ set(LLAMA_CUDA_PEER_MAX_BATCH_SIZE "128" CACHE STRING
"llama: max. batch size for using peer access")
option(LLAMA_CUDA_NO_PEER_COPY "llama: do not use peer to peer copies" OFF)
option(LLAMA_CUDA_NO_VMM "llama: do not try to use CUDA VMM" OFF)
option(LLAMA_CUDA_FA_ALL_QUANTS "llama: compile all quants for FlashAttention" OFF)
option(LLAMA_CURL "llama: use libcurl to download model from an URL" OFF)
option(LLAMA_HIPBLAS "llama: use hipBLAS" OFF)
@ -427,6 +428,9 @@ if (LLAMA_CUDA)
if (LLAMA_CUDA_NO_PEER_COPY)
add_compile_definitions(GGML_CUDA_NO_PEER_COPY)
endif()
if (LLAMA_CUDA_FA_ALL_QUANTS)
add_compile_definitions(GGML_CUDA_FA_ALL_QUANTS)
endif()
if (LLAMA_STATIC)
if (WIN32)