Stretch ICRA Run B090 clean
Private research checkpoint for the Hello Robot Stretch cup-serving policy experiments.
- Experiment arm: Run B090 clean
- Base architecture: NVIDIA GR00T N1.6-3B
- Training data: GT60 + pseudo56; global corr >= 0.90 and global NMAE <= 5% (116 episodes)
- Final training step: 87,125
- Cameras: head + gripper
- Action horizon: 16
Clean retraining from the base model with consistent preprocessing.
Files
This repository contains policy inference weights and the matching processor/statistics/configuration files. DeepSpeed optimizer and RNG states are intentionally excluded.
Use the included statistics.json (or processor/statistics.json, depending on the archived checkpoint layout) and processor configuration with the weights. The exact training configuration is retained under experiment_cfg/.
Intended use
Research and reproducibility only. Validate robot limits, preprocessing, normalization, and emergency-stop behavior before physical deployment.
- Downloads last month
- 8