This repository was archived by the owner on Jun 3, 2025. It is now read-only.
[Text Generation] Avoid mutating the original logits in-place when mapping from logits to token.#1414
Merged
Merged
Conversation
mgoin
reviewed
Nov 17, 2023
mgoin
approved these changes
Nov 20, 2023
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 subscribe to this conversation on GitHub.
Already have an account?
Sign in.
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Feature Description
TokenGenerator.generate(logits: numpy.ndarray)takes a logits array and, ifdo_sample=True, optionally mutates them to enforce the appropriate sampling strategy.This results in the correct generation of tokens, but an in-place modification of the logits array. Logits array is then returned to the user in the mutated form. This can be confusing for the users who are interested in logits value, specially when returned together with prompt logits:
First column
before: the generated logits returned to the user are mutated (the distribution is spikier, arguably close to Dirac distribution)Second column
after: the generated logits are the non-mutatated, original logits, consistent with the prompt logits.