EE26 โ East End Effectors: C1 Insertion Policies
Trained policies for the C1 peg-insertion task on a Franka Emika Panda
(Intel Industrial Robotics Arm Challenge). Each subfolder is a final, train-ready
LeRobot policy (20k steps). Policy IO: observation.state = [8] (7 joints + gripper
width), action = [8] (7 joint positions + gripper).
| Folder | Policy | Dataset | Notes |
|---|---|---|---|
pi0/ |
Pi0 | c1_insertion_merged |
ฯ0 flow-matching policy, ~8.9 GB |
smolvla_merged/ |
SmolVLA | c1_insertion_merged |
SmolVLA on the merged dataset |
smolvla/ |
SmolVLA | c1_insertion |
SmolVLA on the original dataset |
Each folder contains model.safetensors, config.json, train_config.json, and
the LeRobot pre/post-processor (normalizer) files.
Load
from lerobot.policies.factory import make_policy # or the policy-specific class
# e.g. SmolVLAPolicy.from_pretrained("NexusDwin/ee26_east_end_effectors", ...)