Fix non-contiguous tensor problem in jagged_index_select #2060
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Outdated suggestions cannot be applied.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.
Summary:
Before this diff,
jagged_index_select
kernels take raw pointers asarguments. This requires the input tensors to be contiguous.
However, the
jagged_index_select
operator did not make sure that thetensors are contiguous before extracting and passing the raw pointers
to the kernels causing the correctness issue. This diff replaces the
raw pointer arguments with PyTorch's
PackedTensorAccessor
whichhandles non-contiguous tensor accesses automatically. For some
tensors that their raw pointers are still being used, the operator
makes sure that the tensors are contiguous before using them.
Differential Revision: D49937274