fix moe params and run error

vllm/model_executor/layers/fused_moe/fused_moe.py

@@ -1655,6 +1655,7 @@ def fused_experts(
     quant_config: FusedMoEQuantConfig | None = None,
     allow_deep_gemm: bool = False,
     allow_cutlass_block_scaled_grouped_gemm: bool = False,
+    use_nn_moe: bool | None = False,
 ) -> torch.Tensor:
     if quant_config is None:
         quant_config = FUSED_MOE_UNQUANTIZED_CONFIG
@@ -1732,6 +1733,7 @@ def fused_experts(
             block_shape=quant_config.block_shape,
             w1_bias=quant_config.w1_bias,
             w2_bias=quant_config.w2_bias,
+            use_nn_moe=use_nn_moe,
         )
 
 

vllm/model_executor/layers/fused_moe/layer.py

@@ -1985,6 +1985,7 @@ class FusedMoE(CustomOp):
                 if do_naive_dispatch_combine
                 else hidden_states,
                 router_logits=router_logits,
+                use_nn_moe=self.use_nn_moe,
             )
 
             if has_separate_shared_experts: