[Perf] use triton moe backend on hopper by default (#44220)

Signed-off-by: zjy0516 <riverclouds.zhu@qq.com>
2026-06-02 15:52:30 +08:00
parent 68dafcca75
commit dcdfe66bfa
1 changed files with 6 additions and 0 deletions
@@ -68,6 +68,12 @@ def _get_priority_backends(moe_config: FusedMoEConfig) -> list[UnquantizedMoeBac
            UnquantizedMoeBackend.BATCHED_TRITON,
        ]

+        # On Hopper (SM90), the FlashInfer unquantized MoE kernels are slower
+        # than Triton, so prefer Triton by default.
+        if current_platform.is_device_capability_family(90):
+            _move_to_back(_AVAILABLE_BACKENDS, UnquantizedMoeBackend.FLASHINFER_TRTLLM)
+            _move_to_back(_AVAILABLE_BACKENDS, UnquantizedMoeBackend.FLASHINFER_CUTLASS)
+
        # HACK: Qwen3.5 has crash with FLASHINFER_CUTLASS BF16 if DEP.
        # Updating the oracle querying logic is out of the scope of this
        # PR. Need to fix the kernel or update structure in follow up.