perf: incremental metrics - #1036
Merged
Merged
Conversation
Replace O(n²) metric recomputation in GenerateOutputsBuilder with incremental update/compute pattern inspired by torchmetrics. Introduces a Metric protocol and concrete classes (RewardMetric, ErrorRateMetric, InputTokensMetric, OutputTokensMetric, EnvMetrics, PassAtKMetric) in metric_utils.py. Each metric handles its own value extraction from RolloutOutput, making the builder a thin coordinator. Removes the compute_pass_at_k helper in favor of using PassAtKMetric directly. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
mikasenghaas
marked this pull request as ready for review
March 19, 2026 12:15
![cursor[bot]](https://avatars.githubusercontent.com/in/1210556?s=60&v=4){kind=small}
There was a problem hiding this comment.
Cursor Bugbot has reviewed your changes and found 1 potential issue.
Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.
Member
|
@mikasenghaas LGTM pending MC resolve |
willccbb
approved these changes
Mar 19, 2026
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Description
Replace O(n²) metric recomputation in GenerateOutputsBuilder with incremental update/compute pattern. Introduces a Metric protocol and concrete classes (RewardMetric, ErrorRateMetric, InputTokensMetric, OutputTokensMetric, EnvMetrics, PassAtKMetric) in
metric_utils.py. Each metric handles its own value extraction from RolloutOutput, making the builder a thin coordinator. Removes the compute_pass_at_k helper in favor of using PassAtKMetric directly.Note
Medium Risk
Refactors core evaluation metadata computation (reward/error/usage/pass@k) to incremental accumulators, which can subtly change aggregation semantics and affect reported metrics. Low security risk but impacts correctness/performance of evaluation reporting.
Overview
Replaces the one-shot
compute_pass_at_khelper and ad-hoc metadata aggregation with an incremental metric system (Metricprotocol plusRewardMetric,ErrorRateMetric,EnvMetrics, token metrics, andPassAtKMetric).GenerateOutputsBuildernow updates these accumulators as outputs arrive and builds metadata in O(1) without rescanning all outputs.Updates result printing to read
pass_at_k/pass_all_kdirectly from saved metadata (instead of recomputing), and rewrites tests to validate the new metric classes, reset behavior, andGenerateOutputsBuilderpass@k integration.Written by Cursor Bugbot for commit 18f14d6. This will update automatically on new commits. Configure here.