act_transfer_cube / README.md
pravsels's picture
Update model card
e387787 verified
|
Raw
History Blame Contribute Delete
1.16 kB
---
library_name: lerobot
tags:
- act
- robotics
- lerobot
- bimanual
---
# act_transfer_cube
Standard LeRobot ACT trained from scratch on SO101 data.
| | |
|---|---|
| **Policy** | ACT (`policy.type=act`) |
| **Dataset** | [villekuosmanen/armnetbench_transfer_cube](https://huggingface.co/datasets/villekuosmanen/armnetbench_transfer_cube) |
| **Task** | `transfer_cube` |
| **Action dim** | 12 (bimanual) |
| **Cameras** | `top`, `left_wrist`, `right_wrist` |
| **Training** | GCloud A100-80GB (alpha-act-a100-3), batch 16, action horizon 30 |
| **Loss @ 10k** | 0.166 |
| **W&B project** | [act_transfer_cube](https://wandb.ai/pravsels/act_transfer_cube) |
| **W&B run** | [va8dy7k8](https://wandb.ai/pravsels/act_transfer_cube/runs/va8dy7k8) |
## Checkpoints
Archived copies:
| Step | Path |
|------|------|
| 010000 | `checkpoints/010000/pretrained_model/` |
| 008000 | `checkpoints/008000/pretrained_model/` |
The latest step (`010000`) is also copied to the **repository root** for direct loading.
## Usage
```python
from lerobot.policies.act.modeling_act import ACTPolicy
policy = ACTPolicy.from_pretrained("pravsels/act_transfer_cube")
```