search

corl-team / CORL

High-quality single-file implementations of SOTA Offline and Offline-to-Online RL algorithms: AWAC, BC, CQL, DT, EDAC, IQL, SAC-N, TD3+BC, LB-SAC, SPOT, Cal-QL, ReBRAC
https://corl-team.github.io/CORL
Apache License 2.0
465 stars 18 forks source link

Minari Integration with CORL #2

Open Howuhh opened 1 year ago

Howuhh commented 1 year ago

There we will track out progress for Minari integration with CORL. Minari is a standard format for offline RL datasets, with popular reference datasets and related utilities, which we believe will replace D4RL in the future by combining most of the existing benchmarks under unified interface and storage. And we want to be prepared! Eventually this will become a CORLv2.

The plan is to add an experimental separate directory with algorithms using Minari and to retraing all algorithms on the D4RL datasets currently (and in the future) re-created in the Minari.

Adapted algorithms:

Retrained algorithms (only Adroit for now):

masonhargrave commented 10 months ago

This is a really good idea! Would love to see this happen.

Howuhh commented 10 months ago

@masonhargrave It's certainly going to happen! However, you need to be patient as we are waiting for a stable release of Minari to continue working on it. Right now it's missing most of the datasets from D4RL, without which it's hard to benchmark.

jamartinh commented 8 months ago

@masonhargrave @Howuhh

In the meantime, to avoid problems with the large dependencies and unnecessary things of D4RL to just use the datasets I have made an almost single file version of d4lr datasets downloader.

d4rllite

It seems, as well, that Minari versions are very close to be ready soon.