polrolnik2/smolvla_roarm_m3_candy_pick_place

SmolVLA finetuned on rtrtwrw/roarm_m3_candy_pick_place

Finetuned from lerobot/smolvla_base with LeRobot 0.6.2.

dataset	rtrtwrw/roarm_m3_candy_pick_place
steps	20000
batch size	64
precision	bf16
seed	1000
checkpoint	last

Camera keys

This dataset names its cameras front and wrist, while SmolVLA expects camera1/camera2. Training used this rename map, and inference needs the same:

{
  "observation.images.front": "observation.images.camera1",
  "observation.images.wrist": "observation.images.camera2"
}

Usage

from lerobot.policies.smolvla.modeling_smolvla import SmolVLAPolicy
policy = SmolVLAPolicy.from_pretrained("polrolnik2/smolvla_roarm_m3_candy_pick_place")