ONNX support possible?

[tvkamara]

Hi, thank you for your great work. I am curious if there are any possibility to export the PyTorch model to ONNX format? I quickly tried to export the encoder part with torch.onnx.export and torch.onnx.dynamo_export but with no luck. I change the inputs from dict to multiple tensors but still received this error:

RuntimeError: outerNode->outputs().size() == node->inputs().size() INTERNAL ASSERT FAILED at "../torch/csrc/jit/passes/dead_code_elimination.cpp":140, please report a bug to PyTorch.

I'd be interested if anyone has been able to convert the model to ONNX or knows that some parts of the model are definitely not supported. Thanks.