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 2, 04:45 UTC
[FlashAttention] Sync FA with upstream (#44065)
Signed-off-by: Matthew Bonanni <mbonanni@redhat.com>
Matthew Bonanni committed
Jun 2, 2026 at 10:15 UTC
ea0d045a05a5bdd0e0a7d8c8ce94f3ce87c84909
1 file changed
+2
-2
cmake/external_projects/vllm_flash_attn.cmake
+2
-2
@@ -31,7 +31,7 @@ endif()
31
32
if(VLLM_FLASH_ATTN_SRC_DIR)
33
FetchContent_Declare(
34
- vllm-flash-attn SOURCE_DIR
34
+ vllm-flash-attn SOURCE_DIR
35
${VLLM_FLASH_ATTN_SRC_DIR}
36
BINARY_DIR ${CMAKE_BINARY_DIR}/vllm-flash-attn
37
)
@@ -39,7 +39,7 @@ else()
39
FetchContent_Declare(
40
vllm-flash-attn
41
GIT_REPOSITORY https://github.com/vllm-project/flash-attention.git
42
- GIT_TAG bce29425653ec0fbc579d329883030e832d15ada
42
+ GIT_TAG dd62dac706b1cf7895bd99b18c6cb7e7e117ee25
43
GIT_PROGRESS TRUE
44
# Don't share the vllm-flash-attn build between build types
45
BINARY_DIR ${CMAKE_BINARY_DIR}/vllm-flash-attn