Skip to content

coverage: the gaussian-policy · PPO row - #18

Merged
tactino merged 1 commit into
mainfrom
figure/gaussian-ppo
Sep 27, 2026
Merged

tactino merged 1 commit into
mainfrom
figure/gaussian-ppo

Conversation

@tactino

@tactino tactino commented Sep 27, 2026

Copy link
Copy Markdown
Member

The fourth row of "What runs on it": gaussian-policy · PPO, from plugrl-server E38 (PlugRL/plugrl-server#87, figure #88). Merge after those, so that the E38 link in the text resolves.

  • Three new clips and stills: gaussian-ppo-{cheetah,hopper,walker}.
  • coverage.json regenerated: 15 cells. The new row has no square cell, and the page already renders that slot empty.
  • coverage-grid.jpg regenerated with the new row. Its label column is widened to fit "gaussian-policy".
  • index.md and index.zh.md: the paragraph now names the baseline row. The note under the figure says why its square cell is empty and that it ends about where CleanRL's own runs do.

No JavaScript or CSS changes: rows come from the data.

plugrl-server E38 ran a Gaussian policy with PPO, every CleanRL default, on
HalfCheetah, Hopper and Walker2d. It learns all three on 3 of 3 seeds and
ends about where CleanRL's own runs do. The figure gains that row, with no
square cell because it was not run there. Each clip comes from the median
seed's final checkpoint, acting with the mean. coverage.json and the grid
still are regenerated, and both home pages' paragraphs updated to match.
@tactino
tactino merged commit 7594462 into main Sep 27, 2026
1 check passed
@tactino
tactino deleted the figure/gaussian-ppo branch September 27, 2026 22:56
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant