Robotics
LeRobot
Safetensors
smolvla
File size: 1,146 Bytes
d5ef92b
 
 
 
c6e9d86
d5ef92b
c6e9d86
f6e8d5a
 
d5ef92b
f6e8d5a
c6e9d86
 
f6e8d5a
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
---
pipeline_tag: robotics
tags:
- lerobot
library_name: lerobot
---

SmolVLA: A vision-language-action model for affordable and efficient robotics

[Paper](https://huggingface.co/papers/2506.01844)

[Code](https://github.com/huggingface/lerobot)

Designed by Hugging Face.

This model has 450M parameters in total.
You can use inside the [LeRobot library](https://github.com/huggingface/lerobot).

Install smolvla extra dependencies:
```bash
pip install -e ".[smolvla]"
```

Example of finetuning the smolvla pretrained model (`smolvla_base`):
```bash
python lerobot/scripts/train.py \
--policy.path=lerobot/smolvla_base \
--dataset.repo_id=danaaubakirova/svla_so100_task1_v3 \
--batch_size=64 \
--steps=200000
```

Example of finetuning the smolvla neural network with pretrained VLM and action expert
intialized from scratch:
```bash
python lerobot/scripts/train.py \
--policy.type=smolvla \
--dataset.repo_id=danaaubakirova/svla_so100_task1_v3 \
--batch_size=64 \
--steps=200000
```

Example of using the smolvla pretrained model outside LeRobot training framework:
```python
policy = SmolVLAPolicy.from_pretrained("lerobot/smolvla_base")
```