A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 15, 13:30 UTC
[Attention] FA3 decode perf improvement - single mma warp group support for head dim 128 (#16864)
Signed-off-by: Lucas Wilkinson <lwilkinson@neuralmagic.com>
Lucas Wilkinson committed
Apr 24, 2025 at 23:12 UTC
41ca7eb49192cbee65986527319c6eb929d6aa7c
1 file changed
+1
-1
cmake/external_projects/vllm_flash_attn.cmake
+1
-1
@@ -38,7 +38,7 @@ else()
38
FetchContent_Declare(
39
vllm-flash-attn
40
GIT_REPOSITORY https://github.com/vllm-project/flash-attention.git
41
- GIT_TAG 0a721daebe4fa7149f06ecf3d3eabeb6dcd0f1fa
41
+ GIT_TAG e93779c59ba4905e56e5c39dc2c1904ada71fa21
42
GIT_PROGRESS TRUE
43
# Don't share the vllm-flash-attn build between build types
44
BINARY_DIR ${CMAKE_BINARY_DIR}/vllm-flash-attn