fix(opd): score teacher logprobs at rollout temperature, not 0#2085
Open
EazyReal wants to merge 2 commits into
Open
fix(opd): score teacher logprobs at rollout temperature, not 0#2085EazyReal wants to merge 2 commits into
EazyReal wants to merge 2 commits into
Conversation
The on-policy-distillation teacher reward_func scored teacher log-probs via SGLang with a hardcoded temperature=0. SGLang computes input_token_logprobs WITH temperature scaling (compute_temp_top_p_normalized_logprobs), and the student log-probs are temperature-scaled by rollout_temperature (get_responses). So when rollout_temperature != 1 the OPD reverse-KL (student - teacher) compares log-probs at different effective temperatures and is biased. Score the teacher at rollout_temperature so both sides of the KL match. No change at the default rollout_temperature=1.0.
EazyReal
force-pushed
the
opd-teacher-temperature
branch
from
945adff to
6c995f4
Compare
Contributor
Author
|
@zhuzilin could you review this one? OPD teacher scoring was using temperature=0 while rollout samples use --rollout-temperature; this makes the teacher logprobs match the actual sampled distribution instead of silently scoring a greedy distribution. |
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
What changed
Score the on-policy-distillation teacher log-probs at
rollout_temperatureinstead of a hardcodedtemperature: 0inslime/rollout/on_policy_distillation.py.Why
The OPD reverse-KL is
student - teacher. The student log-probs are temperature-scaled byrollout_temperature(get_responses), while the teacher was scored by SGLang attemperature: 0— SGLang scalesinput_token_logprobsby the sampling temperature, so the two sides of the KL were at mismatched effective temperatures wheneverrollout_temperature != 1, biasing the distillation signal. The defaultrollout_temperature=1.0path is byte-identical.Validation
The OPD CI test (
tests/test_qwen2.5_0.5B_opd_sglang.py) runs at--rollout-temperature 0.8, exercising the changed path.