README.md

Offline Reinforcement Learning Codebased on the Unstable Baselines

---

Stable Algorithms (Runnable and have good performance):

Unstable Algorithms (Runnable but does not have SOTA performance)

• Batch-Constrained Deep Q-Learning

Under Development Algorithms

---

Quick Start

Install

Please install unstable_baselines firstly. Then it needs to install D4RL under the unstable_baslines/offline_rl.

pip install git+https://github.com/rail-berkeley/d4rl@master#egg=d4rl

To run an algorithm

python3 /path/to/algorithm/main.py /path/to/algorithm/configs/some-config.json args(optional)