micxael / rl-jepa

RL-JEPA

Paper

Reinforcement Learning Using Joint Embedding Predictive Architecture

How to run

// Suggested using a conda environment

conda create -n "xxxxx" python=3.10

pip install -r requirements.txt

python setup.py build && python setup.py install

// model training and inference
python bin/run_vpg_embed configs/template_vpg_rl_jepa.yml
python bin/run_sac_embed configs/template_sac_rl_jepa.yml
python bin/run_ppo_embed configs/template_ppo_rl_jepa.yml

Reference

This code has been adapted from that of Pritz et al., 2021