chore(release): 0.7.0 - #26
Merged
Merged
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Cuts 0.7.0 for the game-over stop (#25).
An episode can now end because the game ended, not only because it ran out of turns or budget.
advanceRolloutgained a fourth stop condition returning'gameOver'— an ordinary loop exit, sofinalizeRecordruns and the attestation is built exactly as for a turn-limit stop. The pre-existingsignalpath could not be used for this: it throws before the record is finalized, destroying the evidence the grade is made of.Gamegained an optionalover(state). Deriving the flag harness-side was rejected after checking: every substrate publishes a terminal concept inevidence().engineState, but under four different names (ALEterminal, Gymnasiumterminated/truncated, stable-retroepisodeDone, 2048gameOver), so a central derivation would guess at field names and fail silently on the next adapter.Measured on ALE Breakout, seed 0,
maxTurns: 300, scripted policy:stoppedBymaxTurnsstopAtGameOver: truegameOverHalf the decisions, identical verdict. The dropped decisions were verified inert rather than assumed: every evidence channel is byte-identical from decision 150 to 300, while 149 to 150 did move the emulator.
Opt-in, so existing callers keep their fixed denominators. The offset is that
stoppedByandgameOverare recorded on every run whether or not the stop is armed —stoppedBy: 'maxTurns'besidegameOver: truecan only come from an episode played past the end, which is now visible in any artifact.Motivation: a consumer measured 163 of 300 decisions (54.3%) played after the game was over, and could not fix it from outside.