README.md

rl-w-data

This project aims to reproduce some currently popular imitation learning & offline reinforcement learning algorithms, and compare them within the same framework.

done & planning

• ValueDICE
• iq-learn
• sqil

Implementation

This project works with Python 3.7, and before starting the implementation, please check and install the necessary packages listed in ./requirements.txt:

pip install -r requirements.txt

Quick start (ensure that ./run_il_script.sh is executable, and within it, you can choose the algorithms that you want to run):

./run_il_script.sh

Reference

• Imitation Learning via Off-policy Distribution Matching [code]

• IQ-Learn: Inverse soft-Q learning fro Imitation [code]

• SQIL: Imitation Learning via Reinforcement Learning with Sparse Rewards

• pytorch-a2c-ppo-acktr-gail

• stable-baseline3