While exporting model input tensor gets changed from Tensor(1, 1, 32, 151) to Tensor: (tensor(1), tensor(1), tensor(32), tensor(151)). How can I stop this behavior?
Example code is simple:
dummy_input = torch.randn([1, 1, 32, 151], device="cuda")
onnx_reader_path = 'onnx_model.onnx'
torch.onnx.export(Model.model, dummy_input, onnx_reader_path, ...)