xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2026-06-03 17:17:04 +08:00
Code Issues Packages Projects Releases Wiki Activity

[Bugfix] Fix missing seq_start_loc in xformers prefill metadata (#12464)

Browse Source 
Signed-off-by: Isotr0py <2037008807@qq.com>
This commit is contained in:
Isotr0py 2025-01-27 15:25:30 +08:00 committed by GitHub
parent 5204ff5c3f
commit 372bf0890b
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 3 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

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

@ -199,6 +199,8 @@ class XFormersMetadata(AttentionMetadata, PagedAttentionMetadata):
# Compute some attn_metadata fields which default to None # Compute some attn_metadata fields which default to None
query_start_loc = (None if self.query_start_loc is None else query_start_loc = (None if self.query_start_loc is None else
self.query_start_loc[:self.num_prefills + 1]) self.query_start_loc[:self.num_prefills + 1])
seq_start_loc = (None if self.seq_start_loc is None else
self.seq_start_loc[:self.num_prefills + 1])
slot_mapping = (None if self.slot_mapping is None else slot_mapping = (None if self.slot_mapping is None else
self.slot_mapping[:self.num_prefill_tokens]) self.slot_mapping[:self.num_prefill_tokens])
seq_lens = (None if self.seq_lens is None else seq_lens = (None if self.seq_lens is None else
@ -225,6 +227,7 @@ class XFormersMetadata(AttentionMetadata, PagedAttentionMetadata):
max_prefill_seq_len=self.max_prefill_seq_len, max_prefill_seq_len=self.max_prefill_seq_len,
max_decode_seq_len=0, max_decode_seq_len=0,
query_start_loc=query_start_loc, query_start_loc=query_start_loc,
seq_start_loc=seq_start_loc,
context_lens_tensor=context_lens_tensor, context_lens_tensor=context_lens_tensor,
block_tables=block_tables, block_tables=block_tables,
use_cuda_graph=False, use_cuda_graph=False,