File size: 1,409 Bytes
619c7a8
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
---
pipeline_tag: robotics
tags:
- robotics
- manipulation
- pick-and-place
- lerobot
library_name: lerobot
---

# PickCube-v1 LeRobot Dataset

Robot manipulation dataset for pick and place tasks in proper LeRobot format.

## Dataset Structure
- **Episodes**: 5
- **Total Frames**: 250
- **Average Episode Length**: 50 frames
- **Task**: Pick up cube and place at target location

## Features
- `observation.state`: Robot joint positions [6D float32]
- `observation.images.main`: RGB camera feed (256x256)
- `action`: Robot control actions [6D float32]
- `language_instruction`: Task description

## Files Structure
```
dataset/
├── data/chunk-000/          # Parquet files per episode
├── meta/                    # Metadata files
├── videos/chunk-000/        # MP4 videos per episode
└── README.md
```

## Usage

```python
from lerobot.common.datasets.lerobot_dataset import LeRobotDataset

dataset = LeRobotDataset("USERNAME/REPO_NAME")
print(f"Loaded {len(dataset)} frames")

# Access sample
sample = dataset[0]
state = sample['observation.state']
action = sample['action'] 
image = sample['observation.images.main']
```

## Training SmolVLA

```bash
python lerobot/scripts/train.py \
    --policy.path=lerobot/smolvla_base \
    --policy.repo_id=USERNAME/model_name \
    --dataset.repo_id=USERNAME/REPO_NAME \
    --batch_size=8 \
    --steps=2000 \
    --policy.device=cuda
```