in

PyTorch implementation of DreamerV2 model-based RL algorithm


Reimplementation of DreamerV2 model-based RL algorithm in PyTorch.

The official DreamerV2 implementation can be found here.

Features

Running the code

Running locally

Install dependencies

pip3 install -r requirements.txt

Get Atari ROMs

pip3 install atari-py==0.2.9
wget -L -nv http://www.atarimania.com/roms/Roms.rar
apt-get install unrar                                   # brew install unar (Mac)
unrar x Roms.rar                                        # unar -D Roms.rar  (Mac)
unzip ROMS.zip
python3 -m atari_py.import_roms ROMS
rm -rf Roms.rar *ROMS.zip ROMS

Run training (debug CPU mode)

python pydreamer/train.py --configs defaults atari debug --env_id Atari-Pong

Run training (full GPU mode)

python pydreamer/train.py --configs defaults atari atari_pong --run_name atari_pong_1

GitHub

View Github


Leave a Reply

Your email address will not be published. Required fields are marked *

This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.

GIPHY App Key not set. Please check settings

Dish TV Users Can Watch IPL Finals, Mens T20 World Cup for Rs 19 Only

[eBook] The Guide for Reducing SaaS Applications Risk for Lean IT Security Teams