Views
No views yet
| Model Type | Model Name | Parameters | Description |
|---|---|---|---|
| VLA Pretrained | Being-H05-2B | 2B | Base vision-language-action model (preview) |
| VLA Specialist | Being-H05-2B_libero | 2B | Post-trained on LIBERO benchmark |
| VLA Specialist | Being-H05-2B_robocasa | 2B | Post-trained on RoboCasa kitchen tasks |
| VLA Generalist | Being-H05-2B_libero_robocasa | 2B | Post-trained on both LIBERO and RoboCasa |
1git clone https://github.com/BeingBeyond/Being-H05.git
2cd Being-H051conda create -n beingh python=3.10
2conda activate beingh1pip install -r requirements.txt
2pip install flash-attn --no-build-isolation1from BeingH.inference.beingh_policy import BeingHPolicy
2
3# Load model
4policy = BeingHPolicy(
5 model_path="/path/to/Being-H05-2B_libero",
6 device="cuda:0"
7)
8
9# Get action from observation
10action = policy.predict(
11 images=images,
12 state=robot_state,
13 instruction="task prompt"
14)1python BeingH/inference/service.py \
2 --model_path /path/to/Being-H05-2B_libero \
3 --port 8000 \
4 --device cuda:01python BeingH/benchmark/libero/eval_libero.py \
2 --model_path /path/to/Being-H05-2B_libero \
3 --suite libero_spatial \
4 --num_episodes 501python BeingH/benchmark/robocasa/eval_robocasa.py \
2 --model_path /path/to/Being-H05-2B_robocasa \
3 --task PnPCounterToCab \
4 --num_episodes 501torchrun --nproc_per_node=8 BeingH/train/train.py \
2 --mllm_path /path/to/InternVL3_5-2B \
3 --expert_path /path/to/Qwen3-0.6B \
4 --resume_from /path/to/Being-H05-2B \
5 --resume_model_only True \
6 --dataset_config_file configs/posttrain/libero/libero_all.yaml \
7 --output_dir /path/to/output \
8 --max_steps 30000 \
9 --save_steps 10000 \
10 --learning_rate 1e-4 \
11 --action_chunk_length 161@article{beingbeyond2026beingh05,
2 title={Being-H0.5: Scaling Human-Centric Robot Learning for Cross-Embodiment Generalization},
3 author={Luo, Hao and Wang, Ye and Zhang, Wanpeng and Zheng, Sipeng and Xi, Ziheng and Xu, Chaoyi and Xu, Haiweng and Yuan, Haoqi and Zhang, Chi and Wang, Yiqing and Feng, Yicheng and Lu, Zongqing},
4 journal={arXiv preprint arXiv:2601.12993},
5 year={2026}
6}1@article{beingbeyond2025beingh0,
2 title={Being-h0: vision-language-action pretraining from large-scale human videos},
3 author={Luo, Hao and Feng, Yicheng and Zhang, Wanpeng and Zheng, Sipeng and Wang, Ye and Yuan, Haoqi and Liu, Jiazheng and Xu, Chaoyi and Jin, Qin and Lu, Zongqing},
4 journal={arXiv preprint arXiv:2507.15597},
5 year={2025}
6}