Dongkkka/Task_000004_Peanut_Pick_Place_lerobot_Intern_VLA-JEPA-CurrentState-QwenTrain-WMOn_bs16_step5000

VLA-JEPA Current-State Qwen-Train WM-On - Task 000004

LeRobot 0.6.1 inference checkpoint at step 5,000, trained on all 35 episodes of Dongkkka/Task_000004_Peanut_Pick_Place_lerobot_Intern revision 05286a17a145234ed80870702f4d9757f00194c3.

• —Batch size: 16
• —State/action: 22 dimensions
• —Cameras: head, left wrist, right wrist
• —Action chunk / executed actions: 7 / 7
• —Qwen trainable (freeze_qwen=false)
• —World model enabled during training
• —Corrected current-state input path
• —Normalization: state MEANSTD, action MINMAX, visual IDENTITY

This repository contains inference files only. Optimizer, scheduler, and RNG state are kept in the local training checkpoint and are not uploaded. Use the target dataset statistics when constructing VLA-JEPA processors in this Cyclo LeRobot build.