pravsels's picture
Add model card with W&B link
9516c14 verified
|
Raw
History Blame Contribute Delete
1.17 kB
---
library_name: gr00t
tags:
- gr00t
- gr00t-n1.7
- robotics
- vla
base_model: nvidia/GR00T-N1.7-3B
---
# groot1.7_transfer_cube
Fine-tuned [GR00T N1.7-3B](https://huggingface.co/nvidia/GR00T-N1.7-3B) on SO101 data (`transfer_cube`).
| | |
|---|---|
| **Base model** | [nvidia/GR00T-N1.7-3B](https://huggingface.co/nvidia/GR00T-N1.7-3B) |
| **Dataset** | [villekuosmanen/armnetbench_transfer_cube](https://huggingface.co/datasets/villekuosmanen/armnetbench_transfer_cube) |
| **Task** | `transfer_cube` |
| **Training** | Isambard-AI GH200, batch 64, action horizon 16 |
| **W&B project** | [groot1.7_transfer_cube](https://wandb.ai/pravsels/groot1.7_transfer_cube) |
| **W&B run** | [4shbof7g](https://wandb.ai/pravsels/groot1.7_transfer_cube/runs/4shbof7g) |
## Checkpoints
| Step | Loss | Path |
|------|------|------|
| 8,000 | 0.0210 | `checkpoints/8000/pretrained_model/` |
| 6,000 | 0.0234 | `checkpoints/6000/pretrained_model/` |
## Usage
Load from a specific step:
```python
from pathlib import Path
# Use checkpoints/<step>/pretrained_model/ with the GR00T inference stack
ckpt = "pravsels/groot1.7_transfer_cube/checkpoints/8000/pretrained_model"
```