[Fix] Fix Nemotron-H Mamba illegal memory access under DP attention with CUDA graph (#34561)

Co-authored-by: Brayden Zhong <b8zhong@uwaterloo.ca>
2026-08-19 14:13:40 -07:00
parent ed12d6827d
commit a6bc0532c9
2 changed files with 10 additions and 3 deletions
@@ -52,8 +52,6 @@ DP_ATTENTION_EP_ARGS = [
    "0.93",
    "--max-prefill-tokens",
    "8192",
-    "--cuda-graph-backend-prefill",
-    "disabled",
 ]

 MTP_ARGS = [