Based on some initial experiments, it looks like the Encoder classes can not be exported to the ONNX format using torch.onnx.export. It would be great to be able to package them into an ONNX graph along with the rest of the model. Has anyone looked at this?
I think the relevant operators are available in the ONNX spec, so I may give it a try in the near future.
Based on some initial experiments, it looks like the Encoder classes can not be exported to the ONNX format using
torch.onnx.export
. It would be great to be able to package them into an ONNX graph along with the rest of the model. Has anyone looked at this?I think the relevant operators are available in the ONNX spec, so I may give it a try in the near future.