Sample Efficient Reinforcement Learning through Learning from Demonstrations in Minecraft

This repository contains the code of our third-placed submission to the NeurIPS 2019: MineRL Competition.

See our full paper for details: https://arxiv.org/abs/2003.06066.

Run

Download the MineRL dataset to ./data.

# Run local training and testing
python run.py

# Run local training and testing (headless)
xvfb-run -a python run.py

Authors

Yanick Schraner and Christian Scheller

Acknowledgement

The IMPALA implementation is based on https://github.com/deepmind/scalable_agent.