Stuck on an issue?

Lightrun Answers was designed to reduce the constant googling that comes with debugging 3rd party libraries. It collects links to all the places you might be looking at while hunting down a tough bug.

And, if you’re still stuck at the end, we’re happy to hop on a call to see how we can help out.

ONNX export failed on ATen operator upsample_trilinear3d?

See original GitHub issue

Thanks for sharing it. I used it in my network and i got the error as follows

/home/john/anaconda3/lib/python3.6/site-packages/torch/onnx/utils.py:418: UserWarning: ONNX export failed on ATen operator upsample_trilinear3d because torch.onnx.symbolic.upsample_trilinear3d does not exist
  .format(op_name, op_name))

I think I am using the layer likes (pytorch 0.5, python 3.6, Ubuntu 16.04)

x = F.upsample(x, None, 2, 'trilinear')

Issue Analytics

State:
Created 5 years ago
Comments:10 (4 by maintainers)

Top GitHub Comments

1reaction

John1231983commented, Jun 5, 2018

Sorry. I was missing. I have run the old code. Now, it has not error but still not print the graph of model. Let wait. I will paste my model and example

1reaction

lanpacommented, Jun 5, 2018

I go through the code in https://github.com/pytorch/pytorch/blob/master/torch/onnx/symbolic.py There is upsample_bilinear2d, but no upsample_trilinear3d. Can you test again with x = F.upsample(x, None, 2, 'bilinear')? If the error goes away, then I think updating onnx will not help.