Introduce finalizeScoringContext function in label scorers for updating hidden states #146
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
|
I believe in our regular RASR meeting we decided to do this a bit differently and keep the logic only internally in the StatefulLabelScorer instead of modifying the interface or search algorithms directly. I can do the re-write on this branch accordingly but until then, the PR shouldn't be merged. |
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Outdated suggestions cannot be applied.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.
The function
finalizeScoringContext()is added to the LabelScorer. TheStatefulOnnxLabelScorerwon't update the hidden state inextendedScoringContext()anymore, but only append the new token to the sequence. The hidden state update is now done in this new function. The default implementation which is used in the other label scorers is just no-op. For future label scorers, one should pay attention to put costly operations which don't affect the hash intofinalizeScoringContext()instead ofextendedScoringContext()if possible.In the search algorithms, the order is now:
The last step is not done at the end of the decode step, but at the beginning of the next decode step such that we save unnecessary calculations at the end of a segment.