Clarify DFT behavior when inverse=True and onesided=True (irfft)#7574
Merged
justinchuby merged 4 commits intoonnx:mainfrom Jan 26, 2026
Merged
Clarify DFT behavior when inverse=True and onesided=True (irfft)#7574justinchuby merged 4 commits intoonnx:mainfrom
justinchuby merged 4 commits intoonnx:mainfrom
Conversation
This was referenced Jan 15, 2026
4231816 to
9e3ed06
Compare
Codecov Report❌ Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## main #7574 +/- ##
==========================================
- Coverage 55.24% 55.24% -0.01%
==========================================
Files 515 515
Lines 32356 32400 +44
Branches 2894 2896 +2
==========================================
+ Hits 17876 17898 +22
- Misses 13690 13713 +23
+ Partials 790 789 -1 ☔ View full report in Codecov by Sentry. |
Member
|
@onnx/sig-operators |
Contributor
There was a problem hiding this comment.
Pull request overview
This PR clarifies the behavior of the DFT operator when both inverse=True and onesided=True are set, which corresponds to the inverse real FFT (IRFFT) operation. Previously, this combination was prohibited, but it's now properly supported with appropriate validation and documentation.
Changes:
- Removes the restriction preventing
onesided=1andinverse=1from being used together - Adds input validation to ensure RFFT requires real input and IRFFT requires complex input
- Updates documentation and output shape descriptions for all three DFT modes: standard DFT, RFFT, and IRFFT
- Adds comprehensive test coverage for IRFFT functionality
Reviewed changes
Copilot reviewed 5 out of 5 changed files in this pull request and generated 4 comments.
Show a summary per file
| File | Description |
|---|---|
| onnx/defs/math/old.cc | Updates DFT v17 operator schema with IRFFT support, input validation, and corrected shape inference |
| onnx/defs/math/defs.cc | Updates DFT v20 operator schema with IRFFT support, input validation, and corrected shape inference |
| onnx/reference/ops/op_dft.py | Implements IRFFT functionality with proper default dft_length calculation |
| onnx/test/shape_inference_test.py | Adds IRFFT shape inference tests and input validation tests, removes invalid test case |
| onnx/backend/test/case/node/dft.py | Adds end-to-end RFFT and IRFFT test cases for both opset 19 and 20 |
Member
|
@simonbyrne could you please check CI errors and help update the shape inference code? |
Contributor
Author
|
Okay, I think I fixed it. I deleted a few tests which were invalid (they were testing the onesided=true case with complex inputs, which doesn't make any sense). I'm not sure if that counts as a breaking change (it will now error at the shape inference stage, instead of returning a nonsensical result) |
justinchuby
reviewed
Jan 22, 2026
justinchuby
approved these changes
Jan 22, 2026
Member
justinchuby
left a comment
There was a problem hiding this comment.
LGTM in general. A few nitpicking comments
74e681f to
98b353f
Compare
Signed-off-by: Simon Byrne <sbyrne@nvidia.com>
Signed-off-by: Simon Byrne <sbyrne@nvidia.com>
Signed-off-by: Simon Byrne <sbyrne@nvidia.com>
98b353f to
ccf022b
Compare
Member
|
@gramalingam for a second review. |
Signed-off-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
justinchuby
pushed a commit
to microsoft/onnxscript
that referenced
this pull request
Jan 23, 2026
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>
gramalingam
reviewed
Jan 24, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Motivation and Context
Fixes #5920.
This adds some clarity around the behavior of DFT inverse operations, and fully specifies the behavior when inverse=True and onesided=True (corresponding to the irfft operation in numpy and pytorch).