The model conversion from ONNX to X2paddle failed on a VGG model (trained on MNIST)

[shiningrain]

Description

When I'm trying to convert an onnx model (which is constructed by myself, trained on MNIST dataset and using VGG architecture), the conversion of X2paddle throws a value as follows:

In transformed code:

    File "./result_m_v-7/paddle_model/x2paddle_code.py", line 60, in forward
        x2paddle_functional_1_concatenate_concat_0 = paddle.concat(x=[x2paddle_functional_1_resizing_resize_ResizeBilinear_0, x2paddle_functional_1_resizing_resize_ResizeBilinear_0, x2paddle_functional_1_resizing_resize_ResizeBilinear_0], axis=3)
        x2paddle_adjusted_input12 = paddle.transpose(x=x2paddle_functional_1_concatenate_concat_0, perm=[0, 3, 1, 2])
        x2paddle_convolution_output12 = self.conv0(x2paddle_adjusted_input12)
        ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ <--- HERE
        x2paddle_functional_1_vgg16_block1_conv1_Relu_0 = self.relu0(x2paddle_convolution_output12)
        x2paddle_convolution_output11 = self.conv1(x2paddle_functional_1_vgg16_block1_conv1_Relu_0)

    File "./python3.6/site-packages/paddle/fluid/dygraph/layers.py", line 917, in __call__
    return self._dygraph_call_func(*inputs, **kwargs)
    File ".python3.6/site-packages/paddle/fluid/dygraph/layers.py", line 907, in _dygraph_call_func
    outputs = self.forward(*inputs, **kwargs)
    File "./python3.6/site-packages/paddle/nn/layer/conv.py", line 677, in forward
    use_cudnn=self._use_cudnn)
    File "./python3.6/site-packages/paddle/nn/functional/conv.py", line 148, in _conv_nd
    type=op_type, inputs=inputs, outputs=outputs, attrs=attrs)
    File "./python3.6/site-packages/paddle/fluid/layer_helper.py", line 43, in append_op
    return self.main_program.current_block().append_op(*args, **kwargs)
    File "./python3.6/site-packages/paddle/fluid/framework.py", line 3184, in append_op
    attrs=kwargs.get("attrs", None))
    File "./python3.6/site-packages/paddle/fluid/framework.py", line 2344, in __init__
    self.desc.infer_shape(self.block.desc)

    ValueError: (InvalidArgument) The number of input's channels should be equal to filter's channels * groups for Op(Conv). But received: the input's channels is -1, the input's shape is [-1, -1, -1, -1]; the filter's channels is 3, the filter's shape is [64, 3, 3, 3]; the groups is 1, the data_format is NCHW. The error may come from wrong data_format setting.
  [Hint: Expected input_channels == filter_dims[1] * groups, but received input_channels:-1 != filter_dims[1] * groups:3.] (at /paddle/paddle/fluid/operators/conv_op.cc:119)
  [operator < conv2d > error]

This onnx model can be converted to pytorch with onnx2torch but failed when converting to paddle with X2paddle. Could anyone kindly tell me the reason?

Versions

onnx 1.11.0 paddlepaddle 2.2.2 x2paddle 1.4.0

Reproduce

Model

I have put the model (which is about 50 MB) in this repo and you could download it and unzip this file mnist+vgg.zip to get the whole model.

I also uploaded the .zip file of the conversion result in this repo. The paddle_model dir is generated by X2paddle and this ValueError caused the paddle_model/inference_model directory to be empty.

Command

I try to covert this onnx model with the following command and then face the above ValueError.

x2paddle --framework=onnx --model=./.onnx --save_dir=./tmp_pb