[Bug] add flashinfer bool check for fusedmoe in Qwen moe models (#7723)

python/sglang/srt/models/qwen2_moe.py

@@ -143,6 +143,15 @@ class Qwen2MoeSparseMoeBlock(nn.Module):
             renormalize=config.norm_topk_prob,
             quant_config=quant_config,
             prefix=add_prefix("experts", prefix),
+            # Additional args for FusedMoE
+            **(
+                dict(
+                    enable_flashinfer_moe=True,
+                    enable_ep_moe=global_server_args_dict["enable_ep_moe"],
+                )
+                if global_server_args_dict["enable_flashinfer_moe"]
+                else {}
+            ),
         )
 
         self.gate = ReplicatedLinear(

python/sglang/srt/models/qwen3_moe.py

@@ -117,6 +117,15 @@ class Qwen3MoeSparseMoeBlock(nn.Module):
                 if global_server_args_dict["enable_deepep_moe"]
                 else {}
             ),
+            # Additional args for FusedMoE
+            **(
+                dict(
+                    enable_flashinfer_moe=True,
+                    enable_ep_moe=global_server_args_dict["enable_ep_moe"],
+                )
+                if global_server_args_dict["enable_flashinfer_moe"]
+                else {}
+            ),
         )
 
         self.gate = ReplicatedLinear(