code check

Author: ckl117

fastdeploy/model_executor/layers/moe/fused_moe_cutlass_backend.py

@@ -94,11 +94,11 @@ def compute_ffn(
         Paddle Cutlass compute Fused MoE.
         """
         if fastdeploy.envs.FD_USE_PHI_MOE_PERMUTE:
-                out = deep_batch_gemm(permute_input, getattr(layer, self.added_weight_attrs[0]), expert_idx_per_token)
-                out = paddle.nn.functional.swiglu(out)
-                ffn_out_without_down_proj_bias = deep_batch_gemm(
-                    out, getattr(layer, self.added_weight_attrs[1]), expert_idx_per_token
-                )
+            out = deep_batch_gemm(permute_input, getattr(layer, self.added_weight_attrs[0]), expert_idx_per_token)
+            out = paddle.nn.functional.swiglu(out)
+            ffn_out_without_down_proj_bias = deep_batch_gemm(
+                out, getattr(layer, self.added_weight_attrs[1]), expert_idx_per_token
+            )
         else:
             ffn_out_without_down_proj_bias = fastdeploy.model_executor.ops.gpu.moe_expert_ffn(
                 permute_input,