Freeze aerial warmup across training loop

docs/CODEX_EXECUTION_LOG.md

@@ -11817,6 +11817,16 @@ Deployment evidence:
   2 process exposing the old hardcoded value `35` was terminated before any
   result row existed; iteration 2 restarted from the identical checkpoint and
   sampling with the frozen `patience=18` contract on CUDA.
+- The first iteration-2 result row then exposed further generic-default drift:
+  bias LR reached `0.066775` because the direct v37 run's warmup bias `0.01`
+  and aerial HSV settings were not represented in the orchestrator. Both exact
+  processes were stopped after one row, and the invalid run was retained as
+  `iteration-002-invalid-warmup` with a reason marker.
+- Added explicit warmup epochs, warmup bias LR and HSV hue/saturation/value to
+  the orchestrator and versioned completion command. Iteration 2 restarted
+  cleanly from the same candidate and checksummed sampling with
+  `warmup_epochs=1`, `warmup_bias_lr=0.01`, HSV `0.01/0.2/0.15`, patience 18
+  and the existing aerial rotation/flip contract.
 - Added those aerial augmentation parameters to the orchestrator CLI and
   training command, preventing later failure-driven checkpoints from silently
   reverting to generic orientation assumptions.

docs/TODO.md

@@ -978,6 +978,7 @@ This file now starts with the current implementation status. Older preparation/b
 - [x] Bind v37 completion to a one-shot JSON command that enters calibration-first evaluation and up to 20 automatic failure-driven CUDA iterations with the frozen aerial augmentation contract.
 - [x] Complete v37 by early stopping at epoch 23, reject its best epoch-5 checkpoint calibration-first, and start closed-loop iteration 2 from checksummed failure-driven sampling.
 - [x] Make orchestrator patience explicit and retain the frozen value `18` in every subsequent CUDA iteration.
+- [x] Preserve v37 warmup and aerial HSV augmentation in every loop iteration; archive and invalidate the pre-epoch-2 run that exposed generic Ultralytics warmup drift.
 - [x] Evaluate the completed v36 YOLO11x checkpoint calibration-first on the rotated v30 holdouts; reject it before opening test/background because the regional calibration gate failed.
 - [ ] Finish and assess the leak-free v37 YOLO11x failure-driven CUDA iteration; open test/background evidence only if every calibration gate passes.