Remove selective_log_softmax and unify with compute_logprobs #464
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
This PR removes the `selective_log_softmax` function and consolidates all log probability computation through the existing `compute_logprobs` function. Changes: - Removed `selective_log_softmax` from src/forge/util/ops.py - Added `align` parameter to `compute_logprobs` to handle both usage patterns: - align=False: for pre-aligned logits (when model was called with input_ids) - align=True: for extracting subset from full-sequence logits - Updated all references to use `compute_logprobs` with appropriate align flag - Added comprehensive docstring explaining both usage patterns with examples - Removed test_selective_log_softmax.py (functionality now covered by compute_logprobs) The consolidation improves code maintainability by having a single function for log probability computation, while the align parameter handles the two common usage patterns in RL training.
meta-cla
bot
added
the
CLA Signed
|
Could you add unit test for |
https://github.com/meta-pytorch/torchforge/blob/main/tests/unit_tests/util/test_compute_logprobs.py |
Jack-Khuu
approved these changes
Oct 20, 2025
…r tests - Renamed test file to better reflect that it tests ops.py functions - Added three new tests for the align parameter in compute_logprobs: - test_align_parameter_false: validates align=False (pre-aligned logits) - test_align_parameter_true: validates align=True (slicing behavior) - test_align_comparison: verifies slicing logic produces correct results
photomz
pushed a commit
to photomz/forge
that referenced
this pull request
Oct 25, 2025
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
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.
This PR removes the
selective_log_softmaxfunction and consolidates all log probability computation through the existingcompute_logprobsfunction.Changes:
selective_log_softmaxfrom src/forge/util/ops.pyalignparameter tocompute_logprobsto handle both usage patterns:compute_logprobswith appropriate align flagThe consolidation improves code maintainability by having a single function for log probability computation, while the align parameter handles the two common usage patterns in RL training.