---
license: apache-2.0
base_model: HuggingFaceTB/SmolVLM-Instruct
tags:
  - lerobot
  - smolvla
  - vla
  - robotics
  - manipulation
  - real-robot
  - so101
  - visuomotor
  - language-conditioned
datasets:
  - ShubhamK32/so101_declutter_v1
pipeline_tag: robotics
---

# SmolVLA — SO-101 Space Decluttering

SmolVLA policy fine-tuned on the [SO-101 Space Decluttering Dataset v1](https://huggingface.co/datasets/ShubhamK32/so101_declutter_v1) for language-conditioned pick-and-place decluttering tasks on a 6-DoF SO-101 robotic arm. Trained using [LeRobot](https://github.com/huggingface/lerobot).

## Training Details

- **Policy:** SmolVLA (Vision-Language-Action)
- **Steps:** 20,000
- **Robot:** SO-101 6-DoF leader-follower
- **Cameras:** Dual-view — fixed top-view + wrist-mounted egocentric
- **Framework:** LeRobot
- **Language conditioning:** Task descriptions passed as natural language instructions

## Dataset

Trained on [ShubhamK32/so101_declutter_v1](https://huggingface.co/datasets/ShubhamK32/so101_declutter_v1) — a multi-view teleoperation dataset with spatial distractors injected to prevent visual shortcut learning.

## Usage
```python
from lerobot.policies.smolvla.modeling_smolvla import SmolVLAPolicy

policy = SmolVLAPolicy.from_pretrained("ShubhamK32/smolvla_so101_declutter")
```

## Camera Views

- `observation.images.topview` — Fixed overhead. Better for unoccluded pick-place tasks.
- `observation.images.wristview` — Egocentric wrist-mounted. Better for overlapping and cluttered scenes.

## Related

- Dataset: [ShubhamK32/so101_declutter_v1](https://huggingface.co/datasets/ShubhamK32/so101_declutter_v1)
- ACT checkpoint: [ShubhamK32/act_so101_declutter](https://huggingface.co/ShubhamK32/act_so101_declutter)