Repository navigation
Conversation
larissakl
marked this pull request as ready for review
July 15, 2026 08:43
SimBe195
requested changes
Jul 16, 2026
SimBe195
approved these changes
Jul 17, 2026
When every scoring context of a group was already in the score cache, the `continue` skipped the rest of the loop body, including the part that builds the score accessors from the cache. All contexts of that group were then returned as std::nullopt, which the search reads as "not ready to score this context yet". `forwardBatch` already returns early on an empty batch, so the check can simply be dropped. This does not currently happen in a normal decode: a scoring context is identified by its label sequence and its time step, and every decode step appends a label and advances the time step, so the contexts of a step are never cached yet. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
This branch has not been deployed
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Adds a
FullContextOnnxLabelScorer, that forwards either the input feature at the current timestep or the full encoder state sequence, together with the complete label history through a single ONNX model. It's structurally analogous toFixedContextOnnxLabelScorer, but the history is never truncated.I moved the
SeqStepScoringContexttoScoringContexts.[hh|cc]again to reuse it for this label scorer.