README.md
1.6 KB · 63 lines · markdown Raw
1 ---
2 base_model: lerobot/smolvla_base
3 datasets: unknown
4 library_name: lerobot
5 license: apache-2.0
6 model_name: smolvla
7 pipeline_tag: robotics
8 tags:
9 - robotics
10 - lerobot
11 - smolvla
12 ---
13
14 # Model Card for smolvla
15
16 <!-- Provide a quick summary of what the model is/does. -->
17
18
19 [SmolVLA](https://huggingface.co/papers/2506.01844) is a compact, efficient vision-language-action model that achieves competitive performance at reduced computational costs and can be deployed on consumer-grade hardware.
20
21
22 This policy has been trained and pushed to the Hub using [LeRobot](https://github.com/huggingface/lerobot).
23 See the full documentation at [LeRobot Docs](https://huggingface.co/docs/lerobot/index).
24
25 ---
26
27 ## How to Get Started with the Model
28
29 For a complete walkthrough, see the [training guide](https://huggingface.co/docs/lerobot/il_robots#train-a-policy).
30 Below is the short version on how to train and run inference/eval:
31
32 ### Train from scratch
33
34 ```bash
35 lerobot-train \
36 --dataset.repo_id=${HF_USER}/<dataset> \
37 --policy.type=act \
38 --output_dir=outputs/train/<desired_policy_repo_id> \
39 --job_name=lerobot_training \
40 --policy.device=cuda \
41 --policy.repo_id=${HF_USER}/<desired_policy_repo_id>
42 --wandb.enable=true
43 ```
44
45 _Writes checkpoints to `outputs/train/<desired_policy_repo_id>/checkpoints/`._
46
47 ### Evaluate the policy/run inference
48
49 ```bash
50 lerobot-record \
51 --robot.type=so100_follower \
52 --dataset.repo_id=<hf_user>/eval_<dataset> \
53 --policy.path=<hf_user>/<desired_policy_repo_id> \
54 --episodes=10
55 ```
56
57 Prefix the dataset repo with **eval\_** and supply `--policy.path` pointing to a local or hub checkpoint.
58
59 ---
60
61 ## Model Details
62
63 - **License:** apache-2.0