Robot manipulation dataset for pick and place tasks, converted for SmolVLA training.
Episodes: 5
Frames: 250
Image size: 256x256
State dim: 6
Action dim: 6
python -m lerobot.scripts.train
--dataset.repo_id=dancher00/pickcube-v1-smolvla
--policy.path=lerobot/smolvla_base
--batch_size=4
--steps=1000
