Downloads · 30 days
11
23% of all-time downloads
Yu-Zhou-Wang/smolvla_so_101_red_foam_ball
smolvla_so_101_red_foam_ball is a robotics model from Yu-Zhou-Wang. Use it for the robotics task on the model card, and read the license before you ship it in a product. It is set up for lerobot. The card lists the license as apache-2.0.
LeRobot SmolVLA fine-tune of lerobot/smolvlabase on Jingyi-Z/sotac episodes 0–20 (Pick up the red foam ball and place it into the container).
Downloads · 30 days
11
23% of all-time downloads
All-time downloads
48
Public
Parameters
450M
907 MB on disk
Likes
0
Public
Click a slice to open those files.
.safetensors907 MB · 100%
How the weights are stored.
BF16447M · 99%
From the Hugging Face model README
LeRobot SmolVLA fine-tune of lerobot/smolvla_base on
Jingyi-Z/sotac episodes 0–20
(Pick up the red foam ball and place it into the container).
Action expert only (train_expert_only=True), vision encoder frozen. Tactile is not used.
This checkpoint is step 60k / 100k. The 100k job was cancelled at ~77k; 60k is the last saved weights.
| Robot | SO-101 follower, 6 joint positions |
| Cameras | training remaps top → camera1, wrist → camera2; dummy third camera (empty_cameras=1) |
| Chunk size | 50 |
| Steps | 60000 (this checkpoint; run was aimed at 100000) |
| Batch | 8 |
from lerobot.policies.smolvla.modeling_smolvla import SmolVLAPolicy
policy = SmolVLAPolicy.from_pretrained("Yu-Zhou-Wang/smolvla_so_101_red_foam_ball")
Real-robot eval (LeRobot 0.6). Pass the language instruction. If lab cameras are named top / wrist, the checkpoint preprocessor already remaps them.
lerobot-record \
--robot.type=so101_follower \
--policy.path=Yu-Zhou-Wang/smolvla_so_101_red_foam_ball \
--dataset.single_task="Pick up the red foam ball and place it into the container" \
--dataset.repo_id=<you>/eval_smolvla_red_foam_ball
First load also pulls the tokenizer HuggingFaceTB/SmolVLM2-500M-Video-Instruct. VLM weights are already in model.safetensors.
Training code lives on Delta at smolvla_so_101. Dataset: Jingyi-Z/sotac.