biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[Attention][MLA] Re-enable FA4 as default MLA prefill backend (#38819)

Browse Source
This commit is contained in:
Matthew Bonanni
2026-04-06 17:51:46 -04:00
committed by GitHub
parent f186cfe75e
commit 9c81f35b1a
1 changed files with 1 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
vllm/config/attention.py
View File

@@ -30,7 +30,7 @@ class AttentionConfig:
use_cudnn_prefill: bool = False
"""Whether to use cudnn prefill."""
use_trtllm_ragged_deepseek_prefill: bool = True
use_trtllm_ragged_deepseek_prefill: bool = False
"""Whether to use TRTLLM ragged deepseek prefill."""
use_trtllm_attention: bool | None = None