policy_IDM_AUG_40000 — Stretch serve policy (GR00T N1.6-3B)

Base = policy_IDM_AUG (368ep, IDM_GT labels) → continued on GT-only originals 40ep for +20,000 steps (GB64), checkpoint-20000. Intermediate stage of the GT re-anchoring series (see policy_IDM_AUG_60000 for +40,000). Legacy preprocessing (shortest-edge 256), absolute actions.

Serving requirements (critical)

• Normalization: mu3 — STRETCH_N16_MULAW="wrist,gripper" (not stored in ckpt)
• After loading, recompute norm anchors: processor.state_action_processor.use_percentiles=True + _compute_normalization_parameters()
• Preprocessing: legacy shortest-edge 256 (NOT letterbox), absolute actions
• Cameras: dual (gripper wrist + head), task text: "serve the cup"