search

jurgisp / memory-maze

Evaluating long-term memory of reinforcement learning algorithms
MIT License
129 stars 13 forks source link

Note in README to use on headless server #3

Closed dosssman closed 1 year ago

dosssman commented 1 year ago

Greetings.

Thanks a lot for making this suite of environment publicly available. I have taken the liberty of jotting down this PR to trick I found in Dreamer-v2 implementation to make it work on a headless server. Otherwise, it throws up a gladLoadGL error. Hope it can be of some use and sorry in case I missed the official instruction somewhere in the repository.

Best regards.

jurgisp commented 1 year ago

Thanks for spotting, that's a great point!

And thanks for making the README contribution, but I prefer to just set it toMUJOCO_GL=egl in the code so it works by default. I will publish the fix as v1.0.1.