xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2026-07-09 04:17:10 +08:00
Code Issues Packages Projects Releases Wiki Activity

[BugFix][CPU] Fix TorchSDPABackendImpl doesn't have use_irope (#21200)

Browse Source 
Signed-off-by: Lucas Wilkinson <lwilkins@redhat.com>
This commit is contained in:
Lucas Wilkinson 2025-07-19 02:18:48 -04:00 committed by GitHub
parent dcc6cfb991
commit 468e2400fe
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 2 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
vllm/v1/worker/gpu_model_runner.py
View File

@ -2668,7 +2668,8 @@ class GPUModelRunner(LoRAModelRunnerMixin):
# TODO: Support other attention modules, e.g., cross-attention # TODO: Support other attention modules, e.g., cross-attention
if attn_module.attn_type == AttentionType.DECODER: if attn_module.attn_type == AttentionType.DECODER:
use_local_attention = (self.attention_chunk_size is not None use_local_attention = (self.attention_chunk_size is not None
and attn_module.impl.use_irope) and getattr(attn_module.impl,
"use_irope", False))
if attn_module.sliding_window is not None: if attn_module.sliding_window is not None:
kv_cache_spec[layer_name] = SlidingWindowSpec( kv_cache_spec[layer_name] = SlidingWindowSpec(
block_size=block_size, block_size=block_size,