@mailwoman/neural-weights-fr-fr

Stage 2 (coarse + venue/street/house_number) Mailwoman neural-classifier weights.

• locale: fr-fr
• corpus: 0.3.0
• training steps: 2200
• hardware: AMD Radeon 780M (gfx1103) bf16 ~14.6 GiB GTT

Per-component F1 targets

⚠ Below per-component F1 targets:

• country F1 = 0.2112 (target ≥0.95)
• region F1 = 0.1883 (target ≥0.95)
• locality F1 = 0.2736 (target ≥0.95)
• postcode F1 = 0.6916 (target ≥0.95)
• venue F1 = 0.3886 (target ≥0.60)
• street F1 = 0.3016 (target ≥0.70)
• house_number F1 = 0.7866 (target ≥0.80)

Eval (golden set)

• entries: 4535
• full-parse exact match: 0.0818
• mean token confidence: 0.8063

Components supported

Stage 2 ships coarse (country / region / locality / dependent_locality / postcode / subregion / cedex) plus fine-grained venue / street / house_number. Token classifier emits 21 BIO labels.

Files

• model.onnx — int8-quantized ONNX model.
• tokenizer.model — SentencePiece unigram tokenizer (matches the corpus version).
• model-card.json — ModelCard with training + eval metadata.

Loader

Loaded at runtime by @mailwoman/neural. This package contains no JS code.