TBD-VLA single-arm base
| Feature | Shape | Description |
|---|---|---|
observation.images.image |
(3, 256, 256) |
RGB image, channels first |
observation.images.image2 |
(3, 256, 256) |
Second RGB image, channels first |
observation.state |
(8,) |
Robot state |
action |
(7,) per timestep |
Robot action |
The model uses one observation timestep and predicts a chunk of 16 actions, with output shape (batch_size, 16, 7).
- Downloads last month
- 5