E41: E39 resumed to FPO++'s 8M steps - fpo-policy · FPO learns square by the bar, two of three seeds - #96
Merged
Conversation
…n declared as decided after E39
…he bar, two of three seeds past it, all three down from their best
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
E39's
fpo-policy· FPO · square cell, resumed from its final checkpoints and run on to FPO++'s own 8M steps (167 iterations), as pre-registered inPROTOCOL.mdbefore the run.By the registered rule the cell learns: over iterations 158-167, seeds 0 and 1 are +0.291 and +0.214 above their start, and seed 2 is +0.153. P1 and V1 hold, and P2 holds. Its grounds were wrong about which seed would cross: seed 1 did and seed 2 fell back.
What the findings flag: every seed came down from its best window over the last thirty to sixty iterations (seed 2 by 0.09). Fifty-episode evaluations of the final checkpoints are 0.82 / 0.64 / 0.54, against the clone's 0.50 and E39's 0.80 / 0.64 / 0.64.
Only files under
experiments/e41-square-fpo-plus-plus-8m/.