Skip to content

Conversation

@simonbyrne
Copy link
Contributor

This fixes the onnxscript export for the irfft function.

Fixes onnx/onnx#5920, and adds support to the changes in onnx/onnx#7574 and microsoft/onnxruntime#27028.

Most of the diff is due to the onnx_opset generated code changes from onnx/onnx#5920. That can be removed if you would prefer.

Signed-off-by: Simon Byrne <sbyrne@nvidia.com>
Signed-off-by: Simon Byrne <sbyrne@nvidia.com>
axis=dimension,
inverse=True,
onesided=False,
onesided=True,
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Is there a way to implement this with how the dft op is defined currently? I think the current implementation has issues with normalization, but I couldn't figure out what went wrong. You expertise is appreciated!

Given the onnx update is not merged and implemented yet, we are not able to incorporate the change to torchlib at the moment as we treat it as UB (unless the current behavior is actually correct in runtimes?).

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

Development

Successfully merging this pull request may close these issues.

Clarify DFT behavior when inverse=True

2 participants