fix: Resolve compilation bug for empty tensors in `aten::select` by gs-olive · Pull Request #1623 · pytorch/TensorRT (original) (raw)

Description

Add function to utilities to validate input dimensions to Shuffle layers, returning the number of zero dimensions in the input (indicative of empty tensors)
Improve input validation and handling of dynamic shapes in aten::select
Fix bug in usage of squeezeDims regarding use_zeros argument
Add optional argument to squeezeDims to swap 0 and -1, as needed per the input specification
Add regression test case to elicit bug
Add test utility to compare shapes of two input tensors

Fixes #1616 [When applied with #1619]
Fixes #1622

[ x ] My code follows the style guidelines of this project (You can use the linters)
[ x ] I have performed a self-review of my own code
[ x ] I have commented my code, particularly in hard-to-understand areas and hacks
[ x ] I have made corresponding changes to the documentation
[ x ] I have added tests to verify my fix or my feature
[ x ] New and existing unit tests pass locally with my changes
[ x ] I have added the relevant labels to my PR in so that relevant reviewers are notified

Add function to utilities to validate input dimensions to Shuffle layers, returning the number of zero dimensions in the input (indicative of empty tensors)
Improve input validation and handling of dynamic shapes in aten::select
Fix bug in usage of squeezeDims regarding use_zeros argument
Add optional argument to squeezeDims to swap 0 and -1, as needed per the input specification
Add regression test case to elicit bug
Add test utility to compare shapes of two input tensors

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

LGTM