Robotics
Transformers
Safetensors
minicpm_robottrack
feature-extraction
vision-language-action
embodied-ai
minicpm
visual-tracking
custom_code
Instructions to use openbmb/MiniCPM-RobotTrack with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use openbmb/MiniCPM-RobotTrack with Transformers:
# Load model directly from transformers import AutoModel model = AutoModel.from_pretrained("openbmb/MiniCPM-RobotTrack", trust_remote_code=True, device_map="auto") - Notebooks
- Google Colab
- Kaggle
Third-party notices
MiniCPM-RobotTrack integrates with the following projects. Their code, model weights, datasets, and assets remain subject to their own licenses.
- MiniCPM: compact language backbone from OpenBMB.
- DINOv3: visual feature encoder from Meta. The model license is reproduced in
licenses/DINOV3_LICENSE.md. - SigLIP: visual feature encoder from Google.
- Habitat-Lab and Habitat-Sim: simulation framework from Meta. The vendored
Habitat-Lab fork is under
third_party/habitat-laband retains its upstream copyright headers and MIT terms. - EVT-Bench tracking extensions: simulator actions, sensors, and measurements
under
evt_bench. - HM3D, MP3D, humanoid, and robot assets are not redistributed. Users must obtain them from their providers and accept the applicable licenses.