Deep Diffusion Policy Gradient (DDiffPG)

This repository provides a PyTorch implementation of DDiffPG.

---

• ⚙️ Installation
  • Install DDiffPG
• 📜 Usage
  • ✏️ Logging
  • 💡 Train with DDiffPG
  • 🔖 Baselines
  • 💾 Saving and Loading

⚙️ Installation

Install DDiffPG

1. Clone the package:

   git clone [email protected]:ddiffpg.git
   cd ddiffpg

2. Create Conda environment and install dependencies:

   conda env create -f environment.yml
   conda activate ddiffpg

3. Install MuJoCo 210.

4. Install D4RL.

5. Install panda-gym

6. Install DDiffPG.

   pip install -e .

📜 Usage

✏️ Logging

We use Weights & Biases (W&B) for logging.

1. Get a W&B account from https://wandb.ai/site

2. Get your API key from https://wandb.ai/authorize

3. set up your account in terminal

   export WANDB_API_KEY=$API Key$

💡 Train with DDiffPG

Run DDiffPG on AntMaze tasks.

python scripts/ddiffpg_main.py algo=ddiffpg_algo env.name=antmaze-v1

🔖 Baselines

Run DIPO baseline

python scripts/baselines_main.py algo=dipo_algo env.name=antmaze-v1

Run SAC baseline

python scripts/baselines_main.py algo=sac_algo env.name=antmaze-v1

💾 Saving and Loading

Checkpoints are automatically saved as W&B Artifacts.