search

AechPro / distrib-rl

Distributable Reinforcement Learning Platform
Apache License 2.0
7 stars 3 forks source link

Update to gym v0.25.0, with env changes #38

Closed some-rando-rl closed 2 years ago

some-rando-rl commented 2 years ago

Tested on LunarLander-v4 and what's now affectionately known as RocketLeague-v0. For the prior I let it run a batch_size sweep.

In spite of the code working, I still need to clean up a few things in this PR, and revisit some naming. Please don't bother with an in-depth review just yet.

benjamincburns commented 2 years ago

For anyone looking on, some-rando-rl is my alt account, so I submitted the above PR - I'm now abandoning that account, so I'll be contributing here from now on under my real name.

Cosing this and will resubmit as two separate PRs. One that removes the Rocket League custom env altogether, and one that either partially solves #47 (scoped strictly for the version range 0.24.x, 0.25.x), or if that's not a low effort change for the moment, I'll just update to v0.25.0.