Pretrained rllib models not loading

[upasana27]

I tried following this guide: https://github.com/HumanCompatibleAI/overcooked-demo/tree/master/server/static/assets/agents/RllibSelfPlay_CrampedRoom to add my pretrained rllib agent and play with it, however that didn't work. Trying to lead the already given example doesn't work either as shown in attached picture. Is there an ray compatibility issue, as I saw when the tutorial was pushed into this repository, the corresponding ray version in human aware rl repository was lower, is that the issue?

[micahcarroll]

I believe that's probably right – we're in the process of updating human-aware-rl and overcooked-ai.

The branches that the overcooked-demo docker images point to are the master branches of both of those (which have been recently updated, e.g. the rllib version), so this repo might not be currently functional. @jyan1999 will likely be getting to this within the next couple of weeks.

If you need this more urgently, feel free to submit a PR and we will review it! You should only use overcooked-ai now (we have merged all relevant human-aware-rl files into it, and will deprecate the human-aware-rl repo)

[upasana27]

Thank you for the reply, I managed to get it to work with overcooked-ai, but it's a bit of a workaround. Should I submit a PR for review?

[micahcarroll]

Yeah, that would be great! Thank you!!