# One-Stop VLA Development Toolbox for Embodied Intelligence [![Paper](https://img.shields.io/badge/Paper-arXiv-b31b1b.svg)](https://arxiv.org/pdf/2510.23511) [![Hugging Face](https://img.shields.io/badge/%F0%9F%A4%97-Hugging%20Face-yellow)](https://huggingface.co/Dexmal) [![Documentation](https://img.shields.io/badge/Docs-Online-success)](https://dexbotic.com/docs/) [![License](https://img.shields.io/badge/License-MIT-blue.svg)](LICENSE) [![Chinese](https://img.shields.io/badge/%E8%AF%AD%E8%A8%80-%E7%AE%80%E4%BD%93%E4%B8%AD%E6%96%87-red.svg)](README.zh-CN.md)

Pretraining · Fine-tuning · Inference · Evaluation
Supporting mainstream policies such as π0, CogACT, OFT, MemVLA, and more

## Introduction **Dexbotic** is a VLA (Vision-Language-Action) development toolbox built on the PyTorch framework, designed to provide a unified and efficient solution for embodied intelligence research. It comes with built-in environment configurations for various mainstream VLA models, allowing users to reproduce, fine-tune, and inference cutting-edge VLA algorithms with simple setup. - **Ready-to-Use VLA Framework**: Centered around VLA models, integrating embodied manipulation and navigation capabilities, supporting multiple cutting-edge algorithms. - **High-Performance Pre-trained Foundation Models**: For mainstream VLA algorithms such as π0 and CogACT, Dexbotic provides multiple optimized pre-trained models. - **Modular Development Architecture**: Adopting a "layered configuration + factory registration + entry dispatch" architecture, users can easily modify configurations, change models, or add tasks by simply modifying experimental scripts. - **Unified Cloud and Local Training**: Fully supports both cloud and local training needs, supporting cloud training platforms such as Alibaba Cloud and Volcano Engine, while also accommodating consumer-grade GPUs for local training. - **Extensive Robot Compatibility**: For mainstream robots such as UR5, Franka, and ALOHA, Dexbotic provides a **unified training data format** and deployment scripts. ![](resources/intro.png) ## 🔥 News - **[2026-09-11]** Added history-aware inference support to [DM05](docs/DM05.md#history-frame-inference) for richer temporal context. - **[2026-09-10]** Upgraded [DM05](docs/DM05.md#inference-performance) inference with a high-performance backend, delivering an approximately **5x model-inference speedup**. - **[2026-09-01]** Dexbotic [DM05](docs/DM05.md) released! See the [technical report](https://www.dexmal.com/blog/dm0.5) for more information. See [libero_dm05.py](playground/benchmarks/libero/libero_dm05.py) and [libero_dm05_lora.py](playground/benchmarks/libero/libero_dm05_lora.py) for LIBERO full SFT and LoRA. - **[2026-08-06]** Added [CogACT LIBERO LoRA SFT](docs/LiberoLora.md#cogact-lora-recipe) support; see the [CogACT LoRA](playground/benchmarks/libero/libero_cogact_lora.py) for details. - **[2026-08-06]** Added [DM0 LIBERO LoRA SFT](docs/LiberoLora.md#dm0-lora-recipe) support; see the [DM0 LoRA](playground/benchmarks/libero/libero_dm0_lora.py) for details. - **[2026-07-29]** Added [PI05 LIBERO LoRA SFT](docs/LiberoLora.md#pi05-lora-recipe) support; see the [PI05 LoRA](playground/benchmarks/libero/libero_pi05_lora.py) for details. - **[2026-07-29]** Added [PI0 LIBERO LoRA SFT](docs/LiberoLora.md#pi0-lora-recipe) support; see the [PI0 LoRA](playground/benchmarks/libero/libero_pi0_lora.py) for details. - **[2026-07-28]** Added [gr00tsonic](playground/example_gr00tsonic_exp.py) support for Unitree G1 SONIC. See the [Unitree G1 SONIC guide](hardware/unitree_sonic/README.md). - **[2026-06-25]** Added [Co-training](dexbotic/exp/hybrid_dm0_exp.py) capability for the DM0 model. See [Hybrid DM0 Co-Training](docs/DM0.md#hybrid-dm0-co-training) for details. - **[2026-06-18]** Released a [DOS-W1 inference tutorial](hardware/docs/dosw1_inference_example.md), covering how to integrate DOS-W1 with Dexbotic. - **[2026-06-18]** Added a [DM0 realtime inference guide](docs/DM0RealtimeInference.md) for the Triton-backed realtime backend, including 5x core inference speedup results. - **[2026-06-12]** Added a unified [v1 inference API](docs/InferenceAPI.md) for consistent VLA/VLM serving across model wrappers. - **[2026-05-15]** Supported [FSDP2 training backend](docs/FSDP2.md) for faster distributed training. - **[2026-05-09]** Supported [Uni-NaVid](playground/example_uninavid_exp.py) with a [guide](docs/Uni-NaVid.md). - **[2026-04-27]** Supported [RLinf as an RL backend](docs/RLinfAsRLBackend.md) for RL post-training. - **[2026-03-30]** Supported [GR00TN1](playground/benchmarks/libero/libero_gr00tn1.py). - **[2026-03-30]** Added [Co-training](dexbotic/exp/hybrid_pi05_exp.py) capability for the Pi05 model. - **[2026-03-30]** Released a tutorial on integrating [XLeRobot](hardware/docs/xlerobot_inference_example.md) with Dexbotic. - **[2026-02-10]** [DM0](docs/DM0.md) released! See the [technical report](https://dexmal.com/DM0_Tech_Report.pdf) for more information. - **[2026-02-10]** Partnership Announcement: We are excited to announce a strategic collaboration with [RLinf](https://github.com/RLinf/RLinf). Together, our teams will advance VLA + RL research and applications. - **[2026-01-15]** Released a tutorial on integrating [SO-101](hardware/docs/so101_inference_example.md) with Dexbotic. - **[2026-01-15]** Supported [GRPO](docs/RL.md) - **[2026-01-15]** Supported [NaVILA](playground/example_navila_exp.py). - **[2026-01-08]** Added [Co-training](dexbotic/exp/hybrid_cogact_exp.py) capability, enabling joint optimization of action experts and LLMs for the CogACT model. - **[2026-01-08]** Released a specialized image compatible with [Blackwell GPUs](#blackwell-gpus). - **[2025-12-29]** Supported [OFT](playground/benchmarks/libero/libero_oft.py) and [Pi0.5](playground/benchmarks/libero/libero_pi05.py) models. - **[2025-10-20]** Dexbotic officially released! Check out the [technical report](https://arxiv.org/pdf/2510.23511) and [official documentation](https://dexbotic.com/docs/) for details. ## Quick Start We strongly recommend using Docker for development or deployment to get the best experience. ### 1. Installation and Environment Setup ```bash # 1. Clone the repository git clone https://github.com/dexmal/dexbotic.git # 2. Start Docker container docker run -it --rm --gpus all --network host \ -v $(pwd)/dexbotic:/dexbotic \ dexmal/dexbotic \ bash # 3. Activate environment and install dependencies cd /dexbotic conda activate dexbotic pip install -e . ``` > **System Requirements**: Ubuntu 20.04/22.04, recommended GPUs: RTX 4090, A100, or H100 (8 GPUs recommended for training, 1 GPU for deployment).

Using on Blackwell GPUs

For users with Blackwell architecture GPUs (e.g., B100, RTX 5090), please use the specialized Docker image `dexmal/dexbotic:c130t28`. ```bash # 1. Start Docker with Blackwell image docker run -it --rm --gpus all --network host \ -v /path/to/dexbotic:/dexbotic \ dexmal/dexbotic:c130t28 \ bash # 2. Activate environment cd /dexbotic pip install -e . ```

### 2. Usage Guide - [Testing and Evaluation](docs/Tutorial.md#evaluation) - [Training with Provided Simulation Data](docs/Tutorial.md#training-a-model-with-provided-data) - [Training with Your Own Data](docs/Tutorial.md#training-a-model-with-your-own-data) ## Benchmark Results The following shows a comparison of evaluation results between models trained with Dexbotic and original models on mainstream simulation environments. **View more detailed evaluation results**: [Benchmark Results](docs/ModelZoo.md#benchmark-results) ### Libero | Model | Average | Libero-Spatial | Libero-Object | Libero-Goal | Libero-10 | | --- | --- | --- | --- | --- | --- | | CogACT | 93.6 | 97.2 | 98.0 | 90.2 | 88.8 | | DB-CogACT | 94.9 | 93.8 | 97.8 | 96.2 | 91.8 | | π0 | 94.2 | 96.8 | 98.8 | 95.8 | 85.2 | | DB-π0 | 93.9 | 97 | 98.2 | 94 | 86.4 | | MemVLA | 96.7 | 98.4 | 98.4 | 96.4 | 93.4 | | DB-MemVLA | 97.0 | 97.2 | 99.2 | 98.4 | 93.2 | | DB-GR00TN1 | 94.8 | 93.0 | 99.6 | 95.2 | 91.4 | ### CALVIN | Model | Average Length | 1 | 2 | 3 | 4 | 5 | | --- | --- | --- | --- | --- | --- | --- | | CogACT | 3.246 | 83.8 | 72.9 | 64 | 55.9 | 48 | | DB-CogACT | 4.063 | 93.5 | 86.7 | 80.3 | 76 | 69.8 | | OFT | 3.472 | 89.1 | 79.4 | 67.4 | 59.8 | 51.5 | | DB-OFT | 3.540 | 92.8 | 80.7 | 69.2 | 60.2 | 51.1 | ### SimplerEnv | Model | Average | Spoon | Carrot | Stack Blocks | Eggplant | | --- | --- | --- | --- | --- | --- | | CogACT | 51.25 | 71.7 | 50.8 | 15 | 67.5 | | DB-CogACT | 69.45 | 87.5 | 65.28 | 29.17 | 95.83 | | OFT | 30.23 | 12.5 | 4.2 | 4.2 | 100 | | DB-OFT | 76.39 | 91.67 | 76.39 | 43.06 | 94.44 | | MemVLA | 71.9 | 75.0 | 75.0 | 37.5 | 100.0 | | DB-MemVLA | 84.4 | 100.0 | 66.7 | 70.8 | 100.0 | ### ManiSkill2 | Model | Average | PickCube | StackCube | PickSingleYCB | PickSingleEGAD | PickClutterYCB | | --- | --- | --- | --- | --- | --- | --- | | CogACT | 40 | 55 | 70 | 30 | 25 | 20 | | DB-CogACT | 58 | 90 | 65 | 65 | 40 | 30 | | OFT | 21 | 40 | 45 | 5 | 5 | 0 | | DB-OFT | 63 | 90 | 75 | 55 | 65 | 30 | | π0 | 66 | 95 | 85 | 55 | 85 | 10 | | DB-π0 | 65 | 95 | 85 | 65 | 50 | 30 | ### RoboTwin2.0 | Model | Average | Adjust Bottle | Grab Roller | Place Empty Cup | Place Phone Stand | | --- | --- | --- | --- | --- | --- | | CogACT | 43.8 | 87 | 72 | 11 | 5 | | DB-CogACT | 58.5 | 99 | 89 | 28 | 18 | ## FAQ

Q: Failed to install Flash-Attention

A: For detailed installation instructions and troubleshooting, please refer to the official documentation at https://github.com/Dao-AILab/flash-attention.

Q: Coverting RLDS/LeRobot to Dexdata

A: We provide a general data conversion guide in [data conversion](docs/Data.md#2-data-conversion). An example of Lerobot data conversion can be found in [convert_lerobot_to_dexdata](script/convert_data/convert_lerobot_to_dexdata.py), and an example for RLDS data conversion is available in [convert_rlds_to_dexdata](script/convert_data/convert_rlds_to_dexdata.py).

Q: Is 5090 supported?

A: Yes, please refer to [Using on Blackwell GPUs](#blackwell-gpus).

## Support Us We are continuously improving, with more features coming soon. If you like this project, please give us a star on GitHub [![GitHub](https://img.shields.io/github/stars/dexmal/dexbotic?color=5B5BD6)](https://github.com/dexmal/dexbotic). Your support is our motivation to keep moving forward! If Dexbotic has been helpful in your research work, please consider citing our technical report: ```bibtex @article{dexbotic, title={Dexbotic: Open-Source Vision-Language-Action Toolbox}, author={Dexbotic Contributors}, journal={arXiv preprint arXiv:2510.23511}, year={2025} } ``` ## License This project is licensed under the [MIT License](LICENSE).