@tduncan / Nvidia-Optimization-Stack-VLLM Mirror · read-only / commits / 3132a933b6
Fork 0

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 16, 00:45 UTC

[Bugfix][Kernel] FA3 Fix - RuntimeError: This flash attention build only supports pack_gqa (for build size reasons). (#12405)

Signed-off-by: Lucas Wilkinson <lwilkinson@neuralmagic.com>

Lucas Wilkinson committed Jan 24, 2025 at 15:20 UTC 3132a933b65d8ed3383e082264c682940d92d803
1 file changed +1 -1
CMakeLists.txt
+1 -1
@@ -576,7 +576,7 @@ else()
576 FetchContent_Declare(
577 vllm-flash-attn
578 GIT_REPOSITORY https://github.com/vllm-project/flash-attention.git
579 - GIT_TAG 0aff05f577e8a10086066a00618609199b25231d
579 + GIT_TAG 9732b0ce005d1e6216864788502d5570004678f5
580 GIT_PROGRESS TRUE
581 # Don't share the vllm-flash-attn build between build types
582 BINARY_DIR ${CMAKE_BINARY_DIR}/vllm-flash-attn