README.md

---
base_model:
- Qwen/Qwen2.5-VL-7B-Instruct
language:
- en
library_name: transformers
license: apache-2.0
pipeline_tag: robotics
arxiv: 2602.03310
tags:
- RDT
- rdt
- RDT 2
- Vision-Language-Action
- Bimanual
- Manipulation
- Zero-shot
- UMI
---

# RDT2-VQ: Vision-Language-Action with Residual VQ Action Tokens

**RDT2-VQ** is an autoregressive Vision-Language-Action (VLA) model adapted from **[Qwen2.5-VL-7B-Instruct](https://huggingface.co/Qwen/Qwen2.5-VL-7B-Instruct)** and trained on large-scale **UMI** bimanual manipulation data. 

It predicts a short-horizon **relative action chunk** (24 steps, 20 dims/step) from binocular wrist-camera RGB and a natural-language instruction. Actions are discretized with a lightweight **Residual VQ (RVQ)** tokenizer, enabling robust zero-shot transfer across **unseen embodiments** for simple, open-vocabulary skills (e.g., pick, place, shake, wipe).

[**Paper**](https://huggingface.co/papers/2602.03310) - [**Home**](https://rdt-robotics.github.io/rdt2/) - [**Github**](https://github.com/thu-ml/RDT2/tree/main?tab=readme-ov-file) - [**Discord**](https://discord.gg/vsZS3zmf9A)

---

## Table of contents

* [Highlights](#highlights)
* [Model details](#model-details)
* [Hardware & software requirements](#hardware--software-requirements)
* [Quickstart (inference)](#quickstart-inference)
* [Precision settings](#precision-settings)
* [Intended uses & limitations](#intended-uses--limitations)
* [Troubleshooting](#troubleshooting)
* [Changelog](#changelog)
* [Citation](#citation)
* [Contact](#contact)

---

## Highlights

* **Zero-shot cross-embodiment**: Demonstrated on Bimanual **UR5e** and **Franka Research 3** setups; designed to generalize further with correct hardware calibration.
* **UMI scale**: Trained on **10k+ hours** from **100+ indoor scenes** of human manipulation with the UMI gripper.
* **Residual VQ action tokenizer**: Compact, stable action codes; open-vocabulary instruction following via Qwen2.5-VL-7B backbone.

---

## Model details

### Architecture

* **Backbone**: Qwen2.5-VL-7B-Instruct (vision-language).
* **Observation**: Two wrist-camera RGB images (right/left), 384×384, JPEG-like statistics.
* **Instruction**: Short imperative text, recommended format **“Verb + Object.”** (e.g., “Pick up the apple.”).

### Action representation (UMI bimanual, per 24-step chunk)

* 20-D per step = right (10) + left (10):

  * pos (x,y,z): 3
  * rot (6D rotation): 6
  * gripper width: 1
* Output tensor shape: **(T=24, D=20)**, relative deltas, `float32`.
* The RVQ tokenizer yields a fixed-length token sequence; see tokenizer card for exact code lengths.

### Tokenizer

* **Tokenizer repo**: [`robotics-diffusion-transformer/RVQActionTokenizer`](https://huggingface.co/robotics-diffusion-transformer/RVQActionTokenizer)
* Use **float32** for the VQ model.
* Provide a **[LinearNormalizer](http://ml.cs.tsinghua.edu.cn/~lingxuan/rdt2/umi_normalizer_wo_downsample_indentity_rot.pt)** for action scaling (UMI convention).

---

## Hardware & software requirements

Approximate **single-GPU** requirements (Qwen2.5-VL-7B-Instruct scale):

| Mode      |     RAM |    VRAM | Example GPU             |
| --------- | ------: | ------: | ----------------------- |
| Inference | ≥ 32 GB | ≥ 16 GB | RTX 4090                |
| LoRA FT   |       – | ≥ 32 GB | A100 40GB               |
| Full FT   |       – | ≥ 80 GB | A100 80GB / H100 / B200 |

> For **deployment on real robots**, follow your platform’s **end-effector + camera** choices and perform **hardware setup & calibration** (camera stand/pose, flange, etc.) before running closed-loop policies.

**Tested OS**: Ubuntu 24.04.

---

## Quickstart (inference)

```python
# Run under repository: https://github.com/thu-ml/RDT2

import torch
from transformers import AutoProcessor, Qwen2_5_VLForConditionalGeneration

from vqvae import MultiVQVAE
from models.normalizer import LinearNormalizer
from utils import batch_predict_action

# assuming using gpu 0
device = "cuda:0"


processor = AutoProcessor.from_pretrained("Qwen/Qwen2.5-VL-7B-Instruct")
model = Qwen2_5_VLForConditionalGeneration.from_pretrained(
    "robotics-diffusion-transformer/RDT2-VQ",
    torch_dtype=torch.bfloat16,
    attn_implementation="flash_attention_2",
    device_map=device
).eval()
vae = MultiVQVAE.from_pretrained("robotics-diffusion-transformer/RVQActionTokenizer").eval()
vae = vae.to(device=device, dtype=torch.float32)

valid_action_id_length = (
    vae.pos_id_len + vae.rot_id_len + vae.grip_id_len
)
# TODO: modify to your own downloaded normalizer path
# download from http://ml.cs.tsinghua.edu.cn/~lingxuan/rdt2/umi_normalizer_wo_downsample_indentity_rot.pt
normalizer = LinearNormalizer.from_pretrained("umi_normalizer_wo_downsample_indentity_rot.pt")  # 

result = batch_predict_action(
    model,
    processor,
    vae,
    normalizer,
    examples=[
        {
            "obs": {
                # NOTE: following the setting of UMI, camera0_rgb for right arm, camera1_rgb for left arm
                "camera0_rgb": ..., # RGB image in np.ndarray of shape (1, 384, 384, 3) with dtype=np.uint8
                "camera1_rgb": ..., # RGB image in np.ndarray of shape (1, 384, 384, 3) with dtype=np.uint8
            },
            "meta": {
                "num_camera": 2
            }
        },
        ...,    # we support batch inference, so you can pass a list of examples
    ],
    valid_action_id_length=valid_action_id_length,
    apply_jpeg_compression=True,
    # Since model is trained with mostly jpeg images, we suggest toggle this on for better formance
    instruction="Pick up the apple."
    # We suggest using Instruction in format "verb + object" with Capitalized First Letter and trailing period 
)

# get the predict action from example 0
action_chunk = result["action_pred"][0] # torch.FloatTensor of shape (24, 20) with dtype=torch.float32
# action_chunk (T, D) with T=24, D=20
#   T=24: our action_chunk predicts the future 0.8s in fps=30, i.e. 24 frames
#   D=20: following the setting of UMI, we predict the action for both arms from right to left
#   - [0-2]: RIGHT ARM end effector position in x, y, z (unit: m)
#   - [3-8]: RIGHT ARM end effector rotation in 6D rotation representation
#   - [9]: RIGHT ARM gripper width (unit: m)
#   - [10-12]: LEFT ARM end effector position in x, y, z (unit: m)
#   - [13-18]: LEFT ARM end effector rotation in 6D rotation representation
#   - [19]: LEFT ARM gripper width (unit: m)

# rescale gripper width from [0, 0.088] to [0, 0.1]
for robot_idx in range(2):
    action_chunk[:, robot_idx * 10 + 9] = action_chunk[:, robot_idx * 10 + 9] / 0.088 * 0.1
```

> For **installation and fine-tuning instructions**, please refer to the official [GitHub repository](https://github.com/thu-ml/RDT2).

---


## Intended uses & limitations

**Intended uses**

* Research in **robot manipulation** and **VLA modeling**.
* Zero-shot or few-shot deployment on bimanual systems following the repo’s **[hardware calibration](https://github.com/thu-ml/RDT2/tree/main?tab=readme-ov-file#1-important-hard-ware-set-up-and-calibration)** steps.

**Limitations**

* Open-world robustness depends on **calibration quality**, camera placement, and gripper specifics.
* Requires correct **normalization** and **RVQ code compatibility**.
* Safety-critical deployment requires **supervision**, interlocks, and conservative velocity/force limits.

**Safety & responsible use**

* Always test in simulation or with **hardware limits** engaged (reduced speed, gravity compensation, E-stop within reach).

---

## Troubleshooting

| Symptom                            | Likely cause   | Suggested fix                                                       |
| ---------------------------------- | -------------- | ------------------------------------------------------------------- |
| Drifting / unstable gripper widths | Scale mismatch | Apply **LinearNormalizer**; rescale widths (\[0,0.088] → \[0,0.1]). |
| Poor instruction following         | Prompt format  | Use “**Verb + Object.**” with capitalization + period.              |
| No improvement after FT            | OOD actions    | Check RVQ bounds & reconstruction error; verify normalization.      |
| Vision brittleness                 | JPEG gap       | Enable `--image_corruption`; ensure 384×384 inputs.                 |

---

## Changelog

* **2025-09**: Initial release of **RDT2-VQ** on Hugging Face.

---

## Citation

```bibtex
@article{rdt2,
  title={RDT2: Exploring the Scaling Limit of UMI Data Towards Zero-Shot Cross-Embodiment Generalization},
  author={Ji, Xuan and others},
  journal={arXiv preprint arXiv:2602.03310},
  year={2026}
}

@software{rdt2_code,
    title={RDT2: Enabling Zero-Shot Cross-Embodiment Generalization by Scaling Up UMI Data},
    author={RDT Team},
    url={https://github.com/thu-ml/RDT2},
    month={September},
    year={2025}
}
```

---

## Contact

* Project page: [https://rdt-robotics.github.io/rdt2/](https://rdt-robotics.github.io/rdt2/)
* Organization: [https://huggingface.co/robotics-diffusion-transformer](https://huggingface.co/robotics-diffusion-transformer)
* Discord: [https://discord.gg/vsZS3zmf9A](https://discord.gg/vsZS3zmf9A)