Wrap indices to be within max range for as_strided + index_select #3902
Conversation
This file contains 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
sjarus
approved these changes
Dec 12, 2024
vivekkhandelwal1
requested changes
Dec 13, 2024
Rebased PR and CI is clean. Thanks! |
Comment on lines
+503
to
+504
| Value index = | ||
| wrapIndicesAroundMax(b, loc, args[0], input, dimInt); |
There was a problem hiding this comment.
Hi @sahas3, is this a valid thing to do? Can you please share any such reference from the PyTorch code?
There was a problem hiding this comment.
Hi @vivekkhandelwal1, this is based on the behavior I observed -- for the index values being produced, wrapping around the max index value produces correct results for the new e2e test added in the PR. I haven't found any PyTorch implementation to support this behavior though.
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Outdated suggestions cannot be applied.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.
Found these issues when trying to lower the ExportedProgram generated from decomposing
nn.GRUop in torch.as_stridedto reproduce the issue. It requires the other ops in the repro to hit the issue.--config=onnx,index_selectop was generated instead ofas_stridedwhich had the same issue as well.notify : @justin-ngo-arm as you were the original author for the Tosa conversions of these ops.