ONNX import: The number of values in the dilations attribute (2) doesn't match the number of dimensions of the convolutional layer (1).

[salvacarrion]

When loading "FasterRCNN-10_simp.onnx" from the onnx repository (and simplified), the input is not required, which is fine.

But if I type Net *net = import_net_from_onnx_file(model_path);, I get this error:

⚠️  Error in layer 2, the number of values in the dilations attribute (2) doesn't match the number of dimensions of the convolutional layer (1). ⚠️

So... if the input size is indeed required, onnx should warn me about it. Else, this function should simply work.