onnx動態分辨率

onnx導出動態分辨率

pytorch1.4.0版本，不挑cuda版本，mxnet tf挑cuda版本

   output_onnx = 'abcd.onnx'

    input_names = ["images"]
    output_names = ["detects"]
    inputs = torch.randn(1, 3, 360, 640)


    dynamic_axes = {'images': {0: 'batch',2:'batch',3:'batch'}, 'detects': {0: 'batch'}}
    # torch.onnx.export(net, inputs, output_onnx, input_names=inputs, output_names=outputs,
    #                   dynamic_axes=dynamic_axes)

    torch_out = torch.onnx._export(net, inputs, output_onnx, export_params=True, verbose=False,
                                   input_names=input_names, output_names=output_names,opset_version=11,dynamic_axes=dynamic_axes)
    print("==> Exporting model to ONNX format at '{}'".format(output_onnx))