[NVIDIA] Add Cutlass MLA backend (#17625)

2025-06-04 12:40:26 +08:00
parent 8d646c2e53
commit 41aa578428
7 changed files with 111 additions and 3 deletions
--- a/vllm/engine/arg_utils.py
+++ b/vllm/engine/arg_utils.py
@@ -1395,6 +1395,7 @@ class EngineArgs:
            "PALLAS_VLLM_V1",
            "TRITON_ATTN_VLLM_V1",
            "TRITON_MLA",
+            "CUTLASS_MLA_VLLM_V1",
            "FLASHMLA",
            "FLASHINFER",
            "FLASHINFER_VLLM_V1",