Skip to content

Latest commit

 

History

History
36 lines (26 loc) · 1.32 KB

README.md

File metadata and controls

36 lines (26 loc) · 1.32 KB
Raw

DreamRL

Building compact, generative models of environments for efficient reinforcement learning.

See our final presentation here.

Usage

Using conda or venv is probably helpful here. Python version must be 3.7. The base requirements.txt seems to not work as of 2020-11-03, probably because some dependencies broke down. To remedy this, try using the environment.yml

To install with the environment.yml:

conda env create -f environment.yml

To activate:

conda activate DreamRL

human_play is a demo for human players to see how well they can perform in openai gym environments. For human_play.py:

usage: human_play.py [-h] [--plot-rewards] [--zoom-level ZOOM_LEVEL]
                     [--timeout TIMEOUT]
                     env_name

See human_play.py for more details.

Unfortunately, vis.py will not work straight away because the weight files (>1 MB) with the correct dimensions from training are missing from model_weights. If someone from the DreamRL team can help out with this, that would be great!

Contributors

Jonathan Lin (project lead), Joey Hejna, Chelsea Chen, Andrew Chen, Michael Huang, Sumer Kohli, Anish Nag, Jihan Yin