xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2026-05-18 15:56:59 +08:00
Code Issues Packages Projects Releases Wiki Activity

fa format

Browse Source 
Signed-off-by: Sage Moore <sage@neuralmagic.com>
This commit is contained in:
Sage Moore 2025-06-02 19:17:50 +00:00
parent d46397661f
commit e34e4411b9
1 changed files with 9 additions and 7 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

16
vllm/v1/attention/backends/flash_attn.py
View File

@ -324,12 +324,14 @@ class FlashAttentionMetadataBuilder:
scheduler_output: "SchedulerOutput") -> bool: scheduler_output: "SchedulerOutput") -> bool:
return False return False
def build_slice(self, req_slice: slice, def build_slice(
token_slice: slice, self,
max_query_len: int, req_slice: slice,
common_prefix_len: int, token_slice: slice,
common_attn_metadata: CommonAttentionMetadata, max_query_len: int,
) -> FlashAttentionMetadata: common_prefix_len: int,
common_attn_metadata: CommonAttentionMetadata,
) -> FlashAttentionMetadata:
num_reqs = req_slice.stop - req_slice.start num_reqs = req_slice.stop - req_slice.start
num_tokens = token_slice.stop - token_slice.start num_tokens = token_slice.stop - token_slice.start
@ -482,7 +484,7 @@ class FlashAttentionMetadataBuilder:
) )
def use_cascade_attention(self, *args, **kwargs) -> bool: def use_cascade_attention(self, *args, **kwargs) -> bool:
return False #use_cascade_attention(*args, **kwargs) return False #use_cascade_attention(*args, **kwargs)
class FlashAttentionImpl(AttentionImpl): class FlashAttentionImpl(AttentionImpl):