Correct served KL calculation for truncated logprobs - #140
Open
SloptimistPrime wants to merge 1 commit into
Open
SloptimistPrime wants to merge 1 commit into
SloptimistPrime wants to merge 1 commit into
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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 old calculation summed only shared top-logprob entries and divided by their count. It could report a negative KL score.
Use shared tokens plus the remaining probability mass, then average per position. The output labels this as coarsened KL, a lower bound rather than full-vocabulary KL. Also reject invalid probabilities and nonpositive strides.
13 tests pass, including known cases, an HTTP comparison and 200 randomized truncations. This fixes the evaluator; it does not establish that one quantization recipe beats another.
I used AI assistance and reviewed and tested the changes.