binomialdist/lerobot-pi05-train-20260509

ALOHA pi0.5 trained checkpoint

This repository contains a pi0.5 / LeRobot policy checkpoint trained for an ALOHA dual-arm manipulation assignment.

Contents

• model.safetensors: trained policy weights
• config.json: policy configuration
• train_config.json: training configuration
• policy_preprocessor.json and processor weights
• policy_postprocessor.json and processor weights

Training summary

• Training run: outputs/train_20260509_025424
• Final checkpoint: step 20000
• Dataset: binomialdist/lerobot-aloha-dataset
• State/action dimension: 14
• Cameras: cam_high, cam_left_wrist, cam_right_wrist
• Action chunk size: 50

Source note

This checkpoint is provided for course verification. See the accompanying project README/report for environment setup, inference server usage, and experiment analysis.