metadata
license: mit
Model Card for Model ID
This is a hosting of Facebook's VJEPA2-AC with a custom handler to be used with Huggingface Endpoints. View the origional repository here: https://github.com/facebookresearch/vjepa2
- Developed by: [Meta]
- Configured by: [Skyler Wiernik]
- Model type: [Video encoder with action decoder]
- License: [MIT]
Usage
Send a request body like
"inputs": {
"video": video_frames,
"current_pose": current_pose_array
}
where video is an array of base64 jpeg frames with the final frame being the goal and current_pose is an array of (x, y, z, rx, ry, rz, gripper).