Upload README.md with huggingface_hub
Browse files
README.md
ADDED
|
@@ -0,0 +1,30 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
pipeline_tag: robotics
|
| 3 |
+
tags:
|
| 4 |
+
- robotics
|
| 5 |
+
- manipulation
|
| 6 |
+
- smolvla
|
| 7 |
+
- lerobot
|
| 8 |
+
library_name: lerobot
|
| 9 |
+
---
|
| 10 |
+
|
| 11 |
+
# PickCube-v1 Dataset
|
| 12 |
+
|
| 13 |
+
Robot manipulation dataset for pick and place tasks, converted for SmolVLA training.
|
| 14 |
+
|
| 15 |
+
## Dataset Info
|
| 16 |
+
- Episodes: 5
|
| 17 |
+
- Frames: 250
|
| 18 |
+
- Image size: 256x256
|
| 19 |
+
- State dim: 6
|
| 20 |
+
- Action dim: 6
|
| 21 |
+
|
| 22 |
+
## Usage
|
| 23 |
+
|
| 24 |
+
```bash
|
| 25 |
+
python -m lerobot.scripts.train \
|
| 26 |
+
--dataset.repo_id=dancher00/pickcube-v1-smolvla \
|
| 27 |
+
--policy.path=lerobot/smolvla_base \
|
| 28 |
+
--batch_size=4 \
|
| 29 |
+
--steps=1000
|
| 30 |
+
```
|