feat(w4a16): SM90 W4A16 MoE path for DeepSeek-V4-Flash + RSF pre-multiply fix

python/sglang/srt/layers/moe/token_dispatcher/standard.py

@@ -90,6 +90,9 @@ def __init__(self, moe_runner_config: MoeRunnerConfig):
         backend = get_moe_runner_backend()
         self.enable_flashinfer_cutlass_moe = backend.is_flashinfer_cutlass()
         self.enable_flashinfer_mxfp4_moe = backend.is_flashinfer_mxfp4()
+        self.enable_flashinfer_cutlass_wmxfp4a16_moe = (
+            backend.is_flashinfer_cutlass_wmxfp4a16()
+        )
         self.enable_flashinfer_trtllm_routed_moe = backend.is_flashinfer_trtllm_routed()
         # Skip local expert mapping when the backend handles EP with global expert IDs:
         # - cutlass / cutedsl / trtllm_routed handle EP internally
@@ -102,6 +105,7 @@ def __init__(self, moe_runner_config: MoeRunnerConfig):
                 self.enable_flashinfer_mxfp4_moe
                 and envs.SGLANG_OPT_MXFP4_SKIP_DISPATCHER_MAPPING.get()
             )
+            or self.enable_flashinfer_cutlass_wmxfp4a16_moe
         )
         self.num_experts = moe_runner_config.num_experts
         self.num_local_experts = moe_runner_config.num_local_experts

python/sglang/srt/layers/moe/utils.py

@@ -75,6 +75,7 @@ class MoeRunnerBackend(Enum):
     FLASHINFER_TRTLLM_ROUTED = "flashinfer_trtllm_routed"
     FLASHINFER_CUTLASS = "flashinfer_cutlass"
     FLASHINFER_MXFP4 = "flashinfer_mxfp4"
+    FLASHINFER_CUTLASS_WMXFP4A16 = "flashinfer_cutlass_wmxfp4a16"
     FLASHINFER_CUTEDSL = "flashinfer_cutedsl"
     CUTLASS = "cutlass"
     MARLIN = "marlin"
@@ -107,6 +108,9 @@ def is_flashinfer_cutedsl(self):
     def is_flashinfer_mxfp4(self):
         return self == MoeRunnerBackend.FLASHINFER_MXFP4
 
+    def is_flashinfer_cutlass_wmxfp4a16(self):
+        return self == MoeRunnerBackend.FLASHINFER_CUTLASS_WMXFP4A16
+
     def is_cutlass(self):
         return self == MoeRunnerBackend.CUTLASS
 

python/sglang/srt/layers/quantization/fp8.py

@@ -263,6 +263,15 @@ def get_quant_method(
                 )
 
                 return Mxfp4FlashinferTrtllmMoEMethod(fp8_method, prefix=prefix)
+            if (
+                self.is_fp4_experts
+                and get_moe_runner_backend().is_flashinfer_cutlass_wmxfp4a16()
+            ):
+                from sglang.srt.layers.quantization.wmxfp4a16_flashinfer_cutlass_moe import (
+                    Wmxfp4A16FlashinferCutlassMoEMethod,
+                )
+
+                return Wmxfp4A16FlashinferCutlassMoEMethod(fp8_method, prefix=prefix)
             return fp8_method
         elif isinstance(layer, RadixAttention):
             return Fp8KVCacheMethod(self)

python/sglang/srt/layers/quantization/mxfp4_flashinfer_trtllm_moe.py

@@ -37,6 +37,7 @@
 if TYPE_CHECKING:
     from sglang.srt.layers.moe.token_dispatcher import CombineInput, DispatchOutput
 
+
 from sglang.srt.environ import envs
 from sglang.srt.utils.common import get_bool_env_var
 
@@ -132,7 +133,6 @@ def __init__(self, fp8_method, prefix: str):
 
     def create_moe_runner(self, layer, moe_runner_config):
         self.moe_runner_config = moe_runner_config
-
         swiglu_limit = moe_runner_config.swiglu_limit
         assert (
             swiglu_limit is not None