A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 14, 05:30 UTC
[Feature] Batch Invariant for R1 TP 8 on Blackwell (#27229)
Signed-off-by: yewentao256 <zhyanwentao@126.com>
Wentao Ye committed
Oct 21, 2025 at 13:25 UTC
86ed77022db254cab1386c211f150d3840af156c
1 file changed
+1
-1
vllm/model_executor/layers/batch_invariant.py
+1
-1
@@ -756,10 +756,10 @@ def override_envs_for_invariance():
756
"FLEX_ATTENTION",
757
"FLASHINFER",
758
"FLASH_ATTN_MLA",
759
+ "FLASHINFER_MLA",
760
"TRITON_MLA",
761
# Not yet supported MLA backends
762
# "FLASHMLA",
762
- # "FLASHINFER_MLA",
763
]
764
if curr_attn_backend not in supported_backends:
765
warning = (