BatchedExecutor Fixed Forking #621
Merged
Conversation
This file contains 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
…he parent and the child sharing exactly the same logits. Since sampling is allowed to modify logits this could lead to issues in sampling (e.g. one conversation is sampled and overwrites logits to be all zero, second conversation is sampled and generates nonsense). Fixed this by setting a "forked" flag, logits are copied if this flag is set. Flag is cleared next time the conversation is prompted so this extra copying only happens once after a fork occurs.
…y own any unmanaged resources so it is not necessary.
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Outdated suggestions cannot be applied.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.
Previously when a conversation was forked this would result in both the parent and the child sharing exactly the same logits.
Since sampling is allowed to modify logits this could lead to issues in sampling. For example the first conversation samplied modifies the logits to apply a logit bias, the second conversation sampled will implicitly have the same bias!
Fixed this by setting a "forked" flag, logits are copied if this flag is set. Flag is cleared next time the conversation is prompted so this extra copying only happens once after a fork occurs.