I got a problem when I convert mxnet to onnx that l meet a crop layer. The crop layer has two input layers.
I see the file mxnet/contrib/onnx/mx2onnx/_op_translations.py the code is:
if num_inputs > 1: h, w = kwargs["out_shape"][-2:]
but the ‘out_shape’ is not find.so I want to get second input layer shape. or I want to get all hidden layers shape. what should i do.