Installation

Install the required dependencies:

pip install -r requirements.txt

Usage

Training on MuJoCo Tasks

python train_mujoco.py --config_path configs/mujoco/halfcheetah-medium-v2.yaml

Training on AntMaze Tasks

python train_ant.py --config_path configs/antmaze/antmaze-large-diverse.yaml

Training on Kitchen Tasks

python train_kt.py --config_path configs/kitchen/kitchen-mixed.yaml

Training on Adroit Tasks

python train_ad_sb.py --yaml_config_path configs/adroit/pen-human.yaml

Configuration

You can use YAML configuration files from the configs/ directory or override parameters via command line arguments. Command line arguments take priority over configuration files.

Examples

# Using config file
python train_mujoco.py --config_path configs/mujoco/walker2d-medium-v2.yaml

# Overriding config parameters
python train_mujoco.py --config_path configs/mujoco/walker2d-medium-v2.yaml --epochs 100 --beta 0.01

# Using command line arguments only
python train_mujoco.py --env_name walker2d-medium-v2 --epochs 100 --beta 0.01

Project Structure

DRIPC/
├── Algorithms/
│   └── DRIPC.py          # Main DRIPC algorithm implementation
├── configs/              # YAML configuration files
│   ├── mujoco/
│   ├── antmaze/
│   ├── kitchen/
│   └── adroit/
├── utils/                # Utility functions
│   ├── buffer.py
│   ├── common.py
│   └── vd4rl_utils.py
├── train_mujoco.py       # Training script for MuJoCo tasks
├── train_ant.py          # Training script for AntMaze tasks
├── train_kt.py           # Training script for Kitchen tasks
└── train_ad_sb.py        # Training script for Adroit tasks