📊 telemetry: per-round baseline loss (validator_baseline_loss_by_round) - #198
Merged
Conversation
The dashboard needs a stable per-cycle baseline, but validator_baseline_loss
is a single unlabeled gauge overwritten each round, so the gateway can only
sample it and a cycle's baseline ends up timing-dependent.
Add validator_baseline_loss_by_round{round_id} labeled exactly like the
round-attribution families. A new best-effort set_baseline_loss(round_id,
value) helper publishes BOTH the unlabeled gauge (backward compat during
rollout) and the labeled one, and registers the round for eviction; the
foreground-eval baseline site calls it in place of the prior inline set.
The labeled family is added to evict_round_series_before so its stale
labelsets prune on the same cutoff as the other per-round series.
Emitting per-round lets the gateway attribute the right baseline to the
right cycle, freeze it after finalize, and naturally exclude a warming-up
validator that has no value for that round (the gateway averages across
validators, dropping 0/absent).
Tests: helper sets both gauges with the round's id; labeled value is stable
per round; eviction test extended to cover the baseline family. Full
telemetry suite green in the stable image.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
isabella618033
approved these changes
Jul 24, 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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
The dashboard needs a stable per-cycle baseline, but validator_baseline_loss is a single unlabeled gauge overwritten each round, so the gateway can only sample it and a cycle's baseline ends up timing-dependent.
Add validator_baseline_loss_by_round{round_id} labeled exactly like the round-attribution families. A new best-effort set_baseline_loss(round_id, value) helper publishes BOTH the unlabeled gauge (backward compat during rollout) and the labeled one, and registers the round for eviction; the foreground-eval baseline site calls it in place of the prior inline set. The labeled family is added to evict_round_series_before so its stale labelsets prune on the same cutoff as the other per-round series.
Emitting per-round lets the gateway attribute the right baseline to the right cycle, freeze it after finalize, and naturally exclude a warming-up validator that has no value for that round (the gateway averages across validators, dropping 0/absent).
Tests: helper sets both gauges with the round's id; labeled value is stable per round; eviction test extended to cover the baseline family. Full telemetry suite green in the stable image.