OGPO-checkpoints

OGPO Pretrained Checkpoints

Pretrained checkpoints for OGPO: One-Step Generative Policy Optimization for Real-Time Robot Control.

Overview

OGPO enables true one-step generation for real-time robotic control via MeanFlow, dispersive regularization, and RL fine-tuning. These checkpoints can be used directly for fine-tuning with PPO.

Checkpoint Structure

pretrained_checkpoints/
├── OGPO_pretrained_gym_checkpoints/
│   ├── gym_improved_meanflow/              # MeanFlow without dispersive loss
│   └── gym_improved_meanflow_dispersive/   # MeanFlow with dispersive loss (recommended)
└── OGPO_pretraining_robomimic_checkpoints/
    ├── w_0p1/                              # dispersive weight = 0.1
    ├── w_0p5/                              # dispersive weight = 0.5 (recommended)
    └── w_0p9/                              # dispersive weight = 0.9

Supported Tasks

Usage

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

# Gym tasks
base_policy_path: hf://pretrained_checkpoints/OGPO_pretrained_gym_checkpoints/gym_improved_meanflow_dispersive/hopper-medium-v2_best.pt

# Robomimic tasks
base_policy_path: hf://pretrained_checkpoints/OGPO_pretraining_robomimic_checkpoints/w_0p5/can/can_w0p5_08_meanflow_dispersive.pt

Citation

@misc{zou2026ogpo,
      title={OGPO: One-Step Generative Policy Optimization for Real-Time Robot Control},
      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/2601.20701v2},
}

License

MIT License