Fix: Predicted spans NOT being used for anchor conditioning when predict_spans=True#100
Merged
Merged
Conversation
predict_spans() mutates batch.anchor_ids / batch.anchor_alignment. Refresh the cached forward args so the ODE uses predicted temporal anchors.
meta-cla
Bot
added
the
CLA Signed
chevalierNoir
approved these changes
May 26, 2026
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Description
This fixes stale anchor conditioning when
predict_spans=True.Currently,
SAMAudio.separate()populatesforward_argsbeforepredict_spans()is executed.If no anchors are initially provided in the
batch,forward_argsretainsNoneor empty tensors foranchor_idsandanchor_alignment.However,
predict_spans()mutates thebatchby adding predicted anchors throughbatch.process_anchors(...), whileforward_argsis never updated afterward.As a result, the inference continues without any anchor conditioning, ignoring the predicted spans entirely.
Change
After automatic span prediction, refresh:
forward_args["anchor_ids"]forward_args["anchor_alignment"]using the updated batch values and the existing reranking repeat helper.
Notes
I believe this is a significant bug, as it considerably affects output reliability, which is a key aspect highlighted in the paper.