P4.3 formal6 SO101: four real-robot candidates

L1: Legacy P1 seed1 80k. F3: Fixed raw-Top P1 seed3 80k. A1: Original ACT seed1 80k. L1_160K: continued Legacy P1 seed1 160k.

See Chinese test guide for four separate commands. Includes losslessly exported inference weights, Train90 statistics and prototypes, DINO/ResNet offline cache, and matching runtime source snapshot. Training optimizer states remain on the training server and are not needed here. No dataset video or robot calibration is included; retain your own matching nn calibration.

Offline wrapper checks are not real-robot validation. Dev15 uses teacher-forced observations. The original 36-trial block is retained, with a separate user-authorized 12-trial 160k supplement. No real robot has been operated by this preparation task.

Downloads last month: -; Downloads are not tracked for this model. How to track

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support