A while ago, I wrote a pull request that implemented the stubbed-out Pettingzoo_shared_value_function example script, including architecture that allowed users to create instances of MAPPO and MAPPO-inspired algorithms.
I’m finishing a writeup now on the project that required that work, so I remembered the PR and it seems like the thread quietly died when the employee that it was assigned to, HassamSheikh, left AnyScale. I saw that the codebase still doesn’t have a working shared critic example for the new API stack, so I was wondering if there was still interest. I’m happy to adjust the code, if desired. There seemed to be some users that wanted one, so I thought I’d follow up. I’m not sure what the procedure is in situations like this one, since the PR exists and is complete, but the assigned maintainer isn’t there anymore.
If your PR is complete but stalled due to the assigned maintainer leaving, the recommended procedure is to leave a comment on the PR to remove the “stale” label and ping the current RLlib maintainers or active contributors for review. This keeps the PR active and brings it to the attention of the team, as Ray’s process expects contributors to follow up if a PR gets stuck or marked stale. You can also reach out on the Ray discussion forum or Slack to request a new reviewer or escalate the review process if needed. See the official contribution guidelines for details.
There is ongoing interest in a working shared critic (MAPPO) example for the new API stack, as indicated by user feedback and comments on your PR. If you are willing to update the code or address any new review comments, mention this in your PR and ask for reassignment to an active maintainer. This approach is supported by both the Ray contribution docs and recent discussions on similar PRs.
Would you like more detail on how to identify and contact current maintainers, or on best practices for keeping your PR active?