Guowei-Zou/DMPO-datasets

DMPO Demonstration Datasets

Pre-processed demonstration datasets for DMPO: Dispersive MeanFlow Policy Optimization.

Overview

This repository contains pre-processed demonstration data for pre-training DMPO policies. Each dataset includes trajectory data and normalization statistics.

Dataset Structure

gym/
├── hopper-medium-v2/
├── walker2d-medium-v2/
├── ant-medium-expert-v2/
├── Humanoid-medium-v3/
├── kitchen-complete-v0/
├── kitchen-mixed-v0/
└── kitchen-partial-v0/

robomimic/
├── lift-img/
├── can-img/
├── square-img/
└── transport-img/

Each task folder contains:

• train.npz - Training trajectories
• normalization.npz - Observation and action normalization statistics

Usage

Use the hf:// prefix in config files to auto-download:

train_dataset_path: hf://gym/hopper-medium-v2/train.npz
normalization_path: hf://gym/hopper-medium-v2/normalization.npz

Data Sources

• Gym tasks: Derived from D4RL datasets
• Robomimic tasks: Derived from Robomimic proficient-human demonstrations

Citation

@misc{zou2026stepenoughdispersivemeanflow,
      title={One Step Is Enough: Dispersive MeanFlow Policy Optimization},
      author={Guowei Zou and Haitao Wang and Hejun Wu and Yukun Qian and Yuhang Wang and Weibing Li},
      year={2026},
      eprint={2601.20701},
      archivePrefix={arXiv},
      primaryClass={cs.RO},
      url={https://arxiv.org/abs/2601.20701},
}

License

MIT License