deal with comments

x2paddle/op_mapper/onnx2paddle/opset9/opset.py

@@ -1651,47 +1651,18 @@ class OpSet9():
                     outputs=[node.name],
                     shape=shape_list)
             else:
+                # flatten + reshape
                 self.paddle_graph.add_layer(
-                    'paddle.shape',
+                    "paddle.flatten",
                     inputs={"input": val_x.name},
-                    outputs=[val_x.name + "_shape"])
-                self.paddle_graph.add_layer(
-                    "paddle.slice",
-                    inputs={"input": val_x.name + "_shape"},
-                    outputs=[val_x.name + "_shape_first"],
-                    axes=[0],
-                    starts=[0],
-                    ends=[axis])
-                self.paddle_graph.add_layer(
-                    'paddle.prod',
-                    inputs={"x": val_x.name + "_shape_first"},
-                    outputs=[val_x.name + "_shape_first"])
-                self.paddle_graph.add_layer(
-                    "paddle.slice",
-                    inputs={"input": val_x.name + "_shape"},
-                    outputs=[val_x.name + "_shape_second"],
-                    axes=[0],
-                    starts=[axis],
-                    ends=[2147483647])
-                self.paddle_graph.add_layer(
-                    'paddle.prod',
-                    inputs={"x": val_x.name + "_shape_second"},
-                    outputs=[val_x.name + "_shape_second"])
-                self.paddle_graph.add_layer(
-                    'paddle.concat',
-                    inputs={
-                        "x": [
-                            val_x.name + "_shape_first",
-                            val_x.name + "_shape_second"
-                        ]
-                    },
-                    outputs=[val_x.name + "_all_shape"],
-                    axis=0)
+                    outputs=[val_x.name + "_flatten"],
+                    start_axis=[0],
+                    stop_axis=[axis])
                 self.paddle_graph.add_layer(
                     'paddle.reshape',
-                    inputs={'x': val_x.name},
+                    inputs={'x': val_x.name + "_flatten"},
                     outputs=[node.name],
-                    shape=val_x.name + "_all_shape")
+                    shape=[0, -1])
 
     @print_mapping_info
     def Gemm(self, node):