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

metal : use F16 precision in FA kernels

Browse source 
ggml-ci
This commit is contained in:
Georgi Gerganov 2024-11-06 15:33:30 +02:00
parent 25e877309a
commit 7facc29d69
No known key found for this signature in database
GPG key ID: BF970631944C16B7
7 changed files with 476 additions and 333 deletions
Download patch file Download diff file Expand all files Collapse all files

1
ggml/CMakeLists.txt
View file

@ -153,6 +153,7 @@ option(GGML_VULKAN_VALIDATE "ggml: enable Vulkan validation"
option(GGML_VULKAN_RUN_TESTS "ggml: run Vulkan tests" OFF)
option(GGML_KOMPUTE "ggml: use Kompute" OFF)
option(GGML_METAL "ggml: use Metal" ${GGML_METAL_DEFAULT})
option(GGML_METAL_FORCE_FATTN_PREC_F16 "ggml: force F16 accumulators for FA kernels" OFF)
option(GGML_METAL_NDEBUG "ggml: disable Metal debugging" OFF)
option(GGML_METAL_SHADER_DEBUG "ggml: compile Metal with -fno-fast-math" OFF)
option(GGML_METAL_EMBED_LIBRARY "ggml: embed Metal library" ${GGML_METAL})