---
library_name: lerobot
tags:
- smolvla
- robotics
- lerobot
base_model: lerobot/smolvla_base
---

# smolvla_block_stack

Fine-tuned [SmolVLA](https://huggingface.co/lerobot/smolvla_base) on SO101 joint-space data.

| | |
|---|---|
| **Base model** | [lerobot/smolvla_base](https://huggingface.co/lerobot/smolvla_base) |
| **Dataset** | [villekuosmanen/armnetbench_block_stack](https://huggingface.co/datasets/villekuosmanen/armnetbench_block_stack) |
| **Checkpoint step** | 10,000 |
| **Training** | Isambard-AI GH200, batch 128 |
| **W&B project** | [smolvla_block_stack](https://wandb.ai/pravsels/smolvla_block_stack) |
| **W&B run** | [t1bchq5c](https://wandb.ai/pravsels/smolvla_block_stack/runs/t1bchq5c) |

## Usage

```python
from lerobot.policies.smolvla.modeling_smolvla import SmolVLAPolicy
policy = SmolVLAPolicy.from_pretrained("pravsels/smolvla_block_stack")
```