SimpleUnet bug fixes#496
Merged
Merged
Conversation
max_pool, preventing NaN propagation into BatchNorm - Fix 1x1 ConvolutionPlan source/target grids in SimpleUNetDown and SimpleUNetUp to match the grid the data actually resides on - Remove unused imports Fix non-contiguous grad_output crash in convolution_plan.py by calling .contiguous() before passing to the C++ transpose convolution backward kernel. Add tests/unit/test_simple_unet.py with smoke tests covering forward pass, backward pass, batched grids, single-layer UNet, and reset_parameters. Signed-off-by: Jonathan Swartz <jonathan@jswartz.info>
swahtz
added
the
core library
Contributor
There was a problem hiding this comment.
Pull request overview
This PR fixes several bugs in the SimpleUNet implementation and adds test coverage for it. The bugs include NaN propagation via -inf values post max-pooling entering BatchNorm, incorrect ConvolutionPlan source/target grid assignments in SimpleUNetDown and SimpleUNetUp, and a crash on non-contiguous grad_output in the convolution backward kernel. It also removes unused imports.
Changes:
fvdb/convolution_plan.py: Added.contiguous()call ongrad_outputbefore passing it to both the regular and transposed C++ convolution backward kernels, fixing a crash on non-contiguous gradient tensors.fvdb/nn/simple_unet.py: ReplacedGridBatch.max_poolwithfvnn.MaxPoolinSimpleUNetDown(which includes-infcleanup), fixedConvolutionPlansource/target grid arguments in bothSimpleUNetDownandSimpleUNetUpto match the grid where data actually resides after pooling, and removed several unused imports.tests/unit/test_simple_unet.py: New test file with smoke tests covering forward pass, backward pass, batched grids, single-layer UNet, andreset_parameters.
Reviewed changes
Copilot reviewed 3 out of 3 changed files in this pull request and generated 1 comment.
| File | Description |
|---|---|
fvdb/convolution_plan.py |
Adds .contiguous() to grad_output in the backward pass to fix crashes with non-contiguous gradient tensors |
fvdb/nn/simple_unet.py |
Uses fvnn.MaxPool for -inf cleanup, fixes ConvolutionPlan grid arguments, and removes unused imports |
tests/unit/test_simple_unet.py |
New smoke/integration tests for SimpleUNet covering key behaviors |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
You can also share your feedback on Copilot code review. Take the survey.
Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com> Signed-off-by: Jonathan Swartz <jonathan@jswartz.info>
harrism
approved these changes
Mar 4, 2026
harrism
left a comment
harrism
left a comment
Contributor
There was a problem hiding this comment.
Nice to have this fixed and tested.
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.
SimpleUNetDown, preventing NaN propagation into BatchNormAdd tests/unit/test_simple_unet.py with smoke tests covering forward pass, backward pass, batched grids, single-layer UNet, and reset_parameters.