jungfrauandClaude Fable 5.1
8c3c538244
CI / ledger-and-tools (push) Successful in 20s
231 frames unseen by every model (core 148), 82 crystals on them. v4-noflip-mask1 stays the best: F1 0.855, mask score 0.798; v3-noflip-mask1 0.848 / 0.793, within noise; the small cluster model now clearly loses on masks (-0.021, won in 0 % of draws). Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Model Registry
Each trained model lives in its own versioned subdirectory:
models/registry/<version-id>/
├── weights.pt ← primary PyTorch checkpoint
├── exports/ ← derived export formats
│ ├── model.onnx
│ ├── model.engine ← TensorRT (hardware-specific, gitignored)
│ └── pruned_weights.pt
└── metadata.json ← version info and metrics
Naming convention
<architecture>-<task>-<variant>-<YYYY-MM-DD>
Examples:
yolo26n-seg-overlap-false-2026-04-12— yolo26n, segmentation, overlap_mask=False, trained 2026-04-12yolo26n-seg-multiscale-2026-03-16— yolo26n, segmentation, multi-scale training
metadata.json schema
| Field | Description |
|---|---|
model_id |
Same as directory name |
family |
Base architecture (e.g. yolo26n) |
task |
segmentation, detection, etc. |
training_date |
ISO date of training run |
export_formats |
List of available formats |
metrics |
mAP, precision, recall, etc. |
Active production model
See models/active/production.json for the currently deployed version.