This repository contains official implementation of paper Mutual Information Regularized Offline Reinforcement Learning.

Before you start, make sure to run

pip install -e .

Apart from this, you'll have to setup your MuJoCo environment and key as well. Please follow D4RL repo and setup the environment accordingly.

You can run MISA experiments using the following command:

python -m experiments.main --env 'walker2d-medium-v2' --logging.output_dir './experiment_output'

This codebase can also log to W&B online visualization platform. To log to W&B, you first need to set your W&B API key environment variable. Alternatively, you could simply run wandb login.

We provide a wandb link as reference to the reproduced results.

The project heavily borrows from this Jax CQL implementation.

This is not an official Sea Limited or Garena Online Private Limited product.