PORelDICE

This is the implementation for our paper Relaxed Stationary Distribution Correction Estimation for Improved Offline Policy Optimization in Jax.

This codebase is built upon IQL and SQL repository.

Installation

$ conda create -c nvidia -n PORelDICE python=3.8 cuda-nvcc=11.3
$ conda activate PORelDICE
$ pip install -r requirements.txt --no-deps

Mujoco

$ ./run_mujoco.sh

Antmaze

$ ./run_antmaze.sh

Kitchen

$ ./run_kitchen.sh

Name		Name	Last commit message	Last commit date
Latest commit History 8 Commits
configs		configs
wrappers		wrappers
.gitignore		.gitignore
LICENSE		LICENSE
README.md		README.md
actor.py		actor.py
common.py		common.py
critic.py		critic.py
dataset_utils.py		dataset_utils.py
evaluation.py		evaluation.py
learner.py		learner.py
policy.py		policy.py
requirements.txt		requirements.txt
run_antmaze.sh		run_antmaze.sh
run_kitchen.sh		run_kitchen.sh
run_mujoco.sh		run_mujoco.sh
sweep.py		sweep.py
train_offline.py		train_offline.py
value_net.py		value_net.py