Fix __setitem__ in TensorContainer #25
Merged
Conversation
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
added 7 commits
September 11, 2025 19:54
The __setitem__ method in TensorContainer had the indexing loop incorrectly indented inside the tuple check, causing it to skip for non-tuple indices. Moved the loop outside the if block to ensure it runs for all index types. Expanded test cases in test_setitem.py to cover basic, advanced, boolean, multi-dimensional, and edge-case indices for more comprehensive validation.
…r type safety - Add IndexType union type to cover all PyTorch indexing patterns - Update __getitem__ and __setitem__ to use IndexType instead of Any - Fix isinstance check in __setitem__ to use original index variable
Add comprehensive tests for TensorDict slice assignments across different devices (CPU to CUDA and vice versa), demonstrating PyTorch's automatic device transfer for slice operations while highlighting stricter requirements for boolean mask assignments. This ensures device consistency and validates expected behavior in cross-device scenarios.
…ype for type safety Simplify __getitem__ and __setitem__ method signatures by replacing multiple overloads with a unified IndexType, improving code maintainability and reducing redundancy while preserving existing functionality. This change builds on prior fixes to indexing logic.
…e assignments There seems to be no reason, why we would need scalar promotion here.
…indexing methods Added detailed Google-style docstrings to `get_number_of_consuming_dims` and `transform_ellipsis_index`. These docstrings clarify the purpose, arguments, returns, and behavior of these critical indexing methods. `transform_ellipsis_index` is particularly important for ensuring consistent ellipsis expansion across tensors within the container, and its documentation now includes a comprehensive example to illustrate its necessity and function. Inline comments were also added to `transform_ellipsis_index` to explain the step-by-step logic.
The `types.EllipsisType` was introduced in Python 3.10. This change ensures that `EllipsisType` is correctly defined and available when running on Python 3.9 environments, preventing import errors.
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.
Fix TensorContainer Indexing Logic and Enhance Type Safety
Summary
This pull request addresses critical indexing bugs in the TensorContainer system and introduces comprehensive type safety improvements. The changes resolve issues with
__setitem__operations, enhance ellipsis handling, and significantly expand test coverage for all PyTorch indexing patterns.Key Changes
Type Safety
IndexTypetype alias covering all PyTorch indexing patternsIndexTypefor consistent type annotationsTensorDictindexing overloadsBug Fixes
__setitem__logic for non-tuple indices causing assignment failuresDocumentation
get_number_of_consuming_dimsandtransform_ellipsis_indexTesting
Backward Compatibility
All changes maintain backward compatibility. The modifications improve correctness and type safety without breaking existing functionality.