xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2026-05-03 00:04:39 +08:00
Code Issues Packages Projects Releases Wiki Activity

[Bugfix][Kernel] Give unique name to BlockSparseFlashAttention (#12040)

Browse Source 
Signed-off-by: Chen Zhang <zhangch99@outlook.com>
This commit is contained in:
Chen Zhang 2025-01-14 23:45:05 +08:00 committed by GitHub
parent 2e0e017610
commit a2d2acb4c8
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
2 changed files with 2 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
vllm/attention/backends/blocksparse_attn.py
View File

@ -89,8 +89,7 @@ class BlocksparseFlashAttentionBackend(AttentionBackend):
@staticmethod @staticmethod
def get_name() -> str: def get_name() -> str:
# For attention layer compatibility return "BLOCK_SPARSE_FLASH_ATTN"
return "FLASH_ATTN"
@staticmethod @staticmethod
def get_impl_cls() -> Type["BlocksparseFlashAttentionImpl"]: def get_impl_cls() -> Type["BlocksparseFlashAttentionImpl"]:

1
vllm/platforms/interface.py
View File

@ -33,6 +33,7 @@ class _Backend(enum.Enum):
HPU_ATTN = enum.auto() HPU_ATTN = enum.auto()
PALLAS = enum.auto() PALLAS = enum.auto()
IPEX = enum.auto() IPEX = enum.auto()
BLOCK_SPARSE_FLASH_ATTN = enum.auto()
NO_ATTENTION = enum.auto() NO_ATTENTION = enum.auto()