Skip to content

bakirtzisg/ComposeX

Repository files navigation

Quick Start Guide

Installation

  1. Create a new conda environment
conda create -n <env_name> python=3.10
conda activate <env_name>
  1. Install the package
pip install -e .

Running experiments

Baseline experiments

python3 train_baseline_sac.py --env_name <env_name>

The list of baseline environments are ['BaselineCompLift-v1', 'BaselineCompStack-v1', 'BaselineCompNut-v1', 'BaselineCompPickPlaceCan-v1']

Compositional RL experiments

python3 train_comp_sac.py --env_name <env_name>

The list of compositional environments are ['CompLift-v1', 'CompStack-v1', 'CompNut-v1', 'CompPickPlaceCan-v1']

Compositional RL with skill reuse/recycle

python3 train_reuse.py --env_name CompStack-v1
python3 train_reuse.py --env_name CompPickPlaceCan-v1 --keep_learning

Visualize a trained compositional policy

python3 eval_comp_sac.py

About

Compositional arm robot learning

Resources

Stars

Watchers

Forks

Packages

No packages published

Languages