---
license: apache-2.0
base_model: DAVIAN-Robotics/pi05_droid_jointpos
pipeline_tag: robotics
tags:
- robotics
- lerobot
- pi05
- imitation-learning
- discodemo
- franka
---

# DiscoDemo-Stage3_SFT-FMB_Round-alpha0

A visuomotor **π0.5** policy for the **FMB-Round** task (insert a cylindrical peg into its hole on the FMB board), fine-tuned on demonstrations generated
by the P-RFCL baseline generator (α = 0: the same reverse-curriculum RL without the diversity reward) of [DiscoDemo](https://davian-robotics.github.io/DiscoDemo/). It acts from camera images and joint states only,
and transfers to the real Franka FR3 robot without real-world fine-tuning.

| Stage | Repository |
|---|---|
| Generator (RL policy) | [`DAVIAN-Robotics/DiscoDemo-Stage1_RL-FMB_Round-alpha0`](https://huggingface.co/DAVIAN-Robotics/DiscoDemo-Stage1_RL-FMB_Round-alpha0) |
| Generated dataset | [`DAVIAN-Robotics/DiscoDemo-Stage2_GenData-FMB_Round-alpha0`](https://huggingface.co/datasets/DAVIAN-Robotics/DiscoDemo-Stage2_GenData-FMB_Round-alpha0) |
| Imitation policy (π0.5) | [`DAVIAN-Robotics/DiscoDemo-Stage3_SFT-FMB_Round-alpha0`](https://huggingface.co/DAVIAN-Robotics/DiscoDemo-Stage3_SFT-FMB_Round-alpha0) (this repository) |

## Training

| | |
|---|---|
| Base model | [`DAVIAN-Robotics/pi05_droid_jointpos`](https://huggingface.co/DAVIAN-Robotics/pi05_droid_jointpos) (π0.5 DROID, joint-position actions) |
| Data | [`DAVIAN-Robotics/DiscoDemo-Stage2_GenData-FMB_Round-alpha0`](https://huggingface.co/datasets/DAVIAN-Robotics/DiscoDemo-Stage2_GenData-FMB_Round-alpha0) |
| Steps | 20,000 |
| Batch size | 16 |
| Action chunk | 15 (relative to the current joint state) |
| Cameras | over-the-shoulder and wrist |

## Usage

The policy is a [LeRobot](https://github.com/huggingface/lerobot) π0.5 checkpoint. Its pre- and post-processing
pipeline uses DiscoDemo processing steps (relative actions, state tokenization), so loading it requires the DiscoDemo
code: https://github.com/DAVIAN-Robotics/DiscoDemo.

## License

Apache-2.0, following the base model ([openpi](https://github.com/Physical-Intelligence/openpi)).

## Citation

```bibtex
@article{park2026discodemo,
  title   = {DiscoDemo: Discovering Efficient and Diverse Robot Demonstrations for Imitation Learning},
  author  = {Park, Minho and Kim, Kinam and Kim, Donghu and Lee, Byungkun and Hwang, Dongyoon and Shin, Yongjae and Hyung, Junha and Lee, Hojoon and Choo, Jaegul},
  journal = {arXiv preprint},
  year    = {2026}
}
```