Model Card for smolvla

SmolVLA is a compact, efficient vision-language-action model that achieves competitive performance at reduced computational costs and can be deployed on consumer-grade hardware.

This policy has been trained and pushed to the Hub using LeRobot. See the full documentation at LeRobot Docs.

How to Get Started with the Model

For a complete walkthrough, see the training guide. Below is the short version on how to train and run inference/eval:

Train from scratch

lerobot-train \
  --dataset.repo_id=${HF_USER}/<dataset> \
  --policy.type=act \
  --output_dir=outputs/train/<desired_policy_repo_id> \
  --job_name=lerobot_training \
  --policy.device=cuda \
  --policy.repo_id=${HF_USER}/<desired_policy_repo_id>
  --wandb.enable=true

Writes checkpoints to outputs/train/<desired_policy_repo_id>/checkpoints/.

Evaluate the policy/run inference

lerobot-record \
  --robot.type=so100_follower \
  --dataset.repo_id=<hf_user>/eval_<dataset> \
  --policy.path=<hf_user>/<desired_policy_repo_id> \
  --episodes=10

Prefix the dataset repo with eval_ and supply --policy.path pointing to a local or hub checkpoint.

Model Details

License: apache-2.0

Comments

This smolvla policy completes the pick and place task but not always consistent. If the object is placed at a different spot each episode (left, right, near, far) the policy picks up the object correctly.
Checked with changing phone camera zoom, night mode one/off and it does completes the task. However when the phone camera is changed from the one it's trained (changes the zoom, and lighting) it does not work consistently.
Solved the jittering problem upto good extent by makeing policy.num_steps = 20 from earlier 10. Reducing policy.n_action_steps = 50 to 30, 20 does make the task completion worse and don't solve the jittering either.