What Makes a Good Diffusion Planner for Decision Making?

· Paper · Code · OpenReview

This repository contains the PyTorch implementation of "What Makes a Good Diffusion Planner for Decision Making?" (ICLR 2025, Spotlight)

🛠️ Setup

Let's start with python 3.9. It's recommend to create a conda env:

Create a new conda environment

conda create -n dv python=3.9 mesalib glew glfw pip=23 setuptools=63.2.0 wheel=0.38.4 protobuf=3.20 -c conda-forge -y
conda activate dv

Install for MuJoCo Simulator and mujoco-py (Important)

Install mujoco following the instruction here.

Alternatively, run the following script for a quick setup:

#!/bin/bash
sudo apt-get update && sudo apt-get install -y wget tar libosmesa6-dev libgl1-mesa-glx libglfw3 patchelf cmake
sudo ln -s /usr/lib/x86_64-linux-gnu/libGL.so.1 /usr/lib/x86_64-linux-gnu/libGL.so
echo $USER_DIR
wget -c "https://mujoco.org/download/mujoco210-linux-x86_64.tar.gz"
mkdir -p /home/$USER_DIR/.mujoco
cp mujoco210-linux-x86_64.tar.gz /home/$USER_DIR/mujoco.tar.gz
rm mujoco210-linux-x86_64.tar.gz
mkdir -p /home/$USER_DIR/.mujoco
tar -zxvf /home/$USER_DIR/mujoco.tar.gz -C /home/$USER_DIR/.mujoco
echo "export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/home/$USER_DIR/.mujoco/mujoco210/bin" >> ~/.bashrc
echo "export MUJOCO_PY_MUJOCO_PATH=/home/$USER_DIR/.mujoco/mujoco210" >> ~/.bashrc

Install Dependencies

pip install -r requirements.txt
pip install -e .

For PyTorch installation, refer to the official PyTorch setup guide to ensure compatibility with your hardware.

💻 Training & Inference

Run Diffusion Veteran

We provide a single script for easy execution of the training and inference tasksets. You can run the following commands to get started with different environments:

bash scripts/DV_antmaze_reimp.sh
bash scripts/DV_kitchen_reimp.sh
bash scripts/DV_maze2d_reimp.sh
bash scripts/DV_mujoco_reimp.sh

Try Different Configurations

To experiment with different configurations, simply modify the settings in scripts_templates/train_template.sh. This file allows you to easily adjust parameters for your experiments. Once you’ve updated the configuration, run the following script:

bash scripts_templates/template.sh

This provides flexibility for trying different setups without needing to manually edit each script.

🙏 Acknowledgements

🎉 Currently, Diffusion-Veteran is offically supported by CleanDiffuser. You are also welcomed to conduct unified ablation studies or standardized comparisons with the baselines in CleanDiffuser. Please see the license for further details.

📚 Citation

If you find our work useful in your research, please consider citing:

@inproceedings{lu2025what,
  title={What Makes a Good Diffusion Planner for Decision Making?},
  author={Haofei Lu and Dongqi Han and Yifei Shen and Dongsheng Li},
  booktitle={The Thirteenth International Conference on Learning Representations},
  year={2025},
  url={https://openreview.net/forum?id=7BQkXXM8Fy}
}

Also consider citing these prior works that helped contribute to this project:

@article{janner2022planning,
  title={Planning with diffusion for flexible behavior synthesis},
  author={Janner, Michael and Du, Yilun and Tenenbaum, Joshua B and Levine, Sergey},
  journal={arXiv preprint arXiv:2205.09991},
  year={2022}
}
@article{ajay2022conditional,
  title={Is conditional generative modeling all you need for decision-making?},
  author={Ajay, Anurag and Du, Yilun and Gupta, Abhi and Tenenbaum, Joshua and Jaakkola, Tommi and Agrawal, Pulkit},
  journal={arXiv preprint arXiv:2211.15657},
  year={2022}
}
@article{dong2024cleandiffuser,
  title={Cleandiffuser: An easy-to-use modularized library for diffusion models in decision making},
  author={Dong, Zibin and Yuan, Yifu and Hao, Jianye and Ni, Fei and Ma, Yi and Li, Pengyi and Zheng, Yan},
  journal={arXiv preprint arXiv:2406.09509},
  year={2024}
}

🏷️ License

Please see the license for more details.