@mailwoman/neural-weights-fr-fr 3.0.0 → 4.0.0

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (5) hide show
  1. package/README.md +10 -10
  2. package/model-card.json +74 -50
  3. package/model.onnx +0 -0
  4. package/package.json +18 -19
  5. package/tokenizer.model +0 -0
package/README.md CHANGED
@@ -4,26 +4,26 @@ Stage 2 (coarse + venue/street/house_number) Mailwoman neural-classifier weights
4
4
 
5
5
  - locale: **fr-fr**
6
6
  - corpus: **0.3.0**
7
- - training steps: **1800**
7
+ - training steps: **2200**
8
8
  - hardware: **AMD Radeon 780M (gfx1103) bf16 ~14.6 GiB GTT**
9
9
 
10
10
  ## Per-component F1 targets
11
11
 
12
12
  **⚠ Below per-component F1 targets:**
13
13
 
14
- - `country` F1 = **0.2796** (target ≥0.95)
15
- - `region` F1 = **0.1759** (target ≥0.95)
16
- - `locality` F1 = **0.2657** (target ≥0.95)
17
- - `postcode` F1 = **0.7554** (target ≥0.95)
18
- - `venue` F1 = **0.3941** (target ≥0.60)
19
- - `street` F1 = **0.2660** (target ≥0.70)
20
- - `house_number` F1 = **0.7835** (target ≥0.80)
14
+ - `country` F1 = **0.2112** (target ≥0.95)
15
+ - `region` F1 = **0.1883** (target ≥0.95)
16
+ - `locality` F1 = **0.2736** (target ≥0.95)
17
+ - `postcode` F1 = **0.6916** (target ≥0.95)
18
+ - `venue` F1 = **0.3886** (target ≥0.60)
19
+ - `street` F1 = **0.3016** (target ≥0.70)
20
+ - `house_number` F1 = **0.7866** (target ≥0.80)
21
21
 
22
22
  ## Eval (golden set)
23
23
 
24
24
  - entries: **4535**
25
- - full-parse exact match: **0.1074**
26
- - mean token confidence: **0.8566**
25
+ - full-parse exact match: **0.0818**
26
+ - mean token confidence: **0.8063**
27
27
 
28
28
  ## Components supported
29
29
 
package/model-card.json CHANGED
@@ -1,17 +1,18 @@
1
1
  {
2
2
  "name": "neural-weights-fr-fr",
3
- "version": "3.0.0",
3
+ "version": "4.0.0",
4
+ "model_lineage": "shares the en-us Stage 3 / step 100000 model (formerly v0.6.0) — relabeled to the unified 4.0.0 release version; tokenizer 0.6.0-a0",
4
5
  "phase": "Stage 2 (coarse + venue/street/house_number)",
5
6
  "license": "AGPL-3.0-only",
6
7
  "locale": "fr-fr",
7
8
  "training": {
8
9
  "corpus_version": "0.3.0",
9
10
  "tokenizer_version": "0.1.0",
10
- "steps": 1800,
11
+ "steps": 2200,
11
12
  "hardware": "AMD Radeon 780M (gfx1103) bf16 ~14.6 GiB GTT",
12
- "duration_seconds": 1067.0,
13
+ "duration_seconds": 1146.0,
13
14
  "started_at": null,
14
- "completed_at": "2026-05-22T10:22:18.739046Z"
15
+ "completed_at": "2026-05-23T06:21:51.190078Z"
15
16
  },
16
17
  "components_supported": [
17
18
  "country",
@@ -25,39 +26,62 @@
25
26
  "street",
26
27
  "house_number"
27
28
  ],
29
+ "labels": [
30
+ "O",
31
+ "B-country",
32
+ "I-country",
33
+ "B-region",
34
+ "I-region",
35
+ "B-locality",
36
+ "I-locality",
37
+ "B-dependent_locality",
38
+ "I-dependent_locality",
39
+ "B-postcode",
40
+ "I-postcode",
41
+ "B-subregion",
42
+ "I-subregion",
43
+ "B-cedex",
44
+ "I-cedex",
45
+ "B-venue",
46
+ "I-venue",
47
+ "B-street",
48
+ "I-street",
49
+ "B-house_number",
50
+ "I-house_number"
51
+ ],
28
52
  "eval": {
29
53
  "n_entries": 4535,
30
- "full_parse_exact_match": 0.1073869900771775,
31
- "mean_token_confidence": 0.856554333787935,
54
+ "full_parse_exact_match": 0.08180815876515987,
55
+ "mean_token_confidence": 0.8062812768727202,
32
56
  "per_component": {
33
57
  "country": {
34
- "precision": 0.2954545454532025,
35
- "recall": 0.2653061224478967,
36
- "f1": 0.2795698919733611,
58
+ "precision": 0.21428571428481394,
59
+ "recall": 0.20816326530527282,
60
+ "f1": 0.21118012372283307,
37
61
  "support": 245
38
62
  },
39
63
  "region": {
40
- "precision": 0.4411027568916778,
41
- "recall": 0.10982839313569116,
42
- "f1": 0.17586809860649805,
64
+ "precision": 0.342951360263526,
65
+ "recall": 0.129797191887635,
66
+ "f1": 0.18832050661831204,
43
67
  "support": 3205
44
68
  },
45
69
  "locality": {
46
- "precision": 0.23303370786511615,
47
- "recall": 0.3089067619897799,
48
- "f1": 0.26565902346259507,
70
+ "precision": 0.24782398452605223,
71
+ "recall": 0.30533214179317686,
72
+ "f1": 0.2735886822759171,
49
73
  "support": 3357
50
74
  },
51
75
  "dependent_locality": {
52
- "precision": 0.0,
53
- "recall": 0.0,
54
- "f1": 0.0,
76
+ "precision": 0.005044136191670815,
77
+ "recall": 0.0999999999975,
78
+ "f1": 0.009603841445164863,
55
79
  "support": 40
56
80
  },
57
81
  "postcode": {
58
- "precision": 0.8426270136303831,
59
- "recall": 0.684563758389032,
60
- "f1": 0.7554156632710455,
82
+ "precision": 0.8323890462696731,
83
+ "recall": 0.591610738254835,
84
+ "f1": 0.6916437813892687,
61
85
  "support": 2980
62
86
  },
63
87
  "subregion": {
@@ -73,21 +97,21 @@
73
97
  "support": 1
74
98
  },
75
99
  "venue": {
76
- "precision": 0.38622493461169466,
77
- "recall": 0.4023614895545846,
78
- "f1": 0.39412811337886233,
100
+ "precision": 0.37649063032335905,
101
+ "recall": 0.4014532243411431,
102
+ "f1": 0.38857142807160183,
79
103
  "support": 1101
80
104
  },
81
105
  "street": {
82
- "precision": 0.3326499231161801,
83
- "recall": 0.22165300546440517,
84
- "f1": 0.2660381220860398,
106
+ "precision": 0.35594795539016916,
107
+ "recall": 0.26161202185783416,
108
+ "f1": 0.3015748026611547,
85
109
  "support": 2928
86
110
  },
87
111
  "house_number": {
88
- "precision": 0.7315737051789185,
89
- "recall": 0.8432835820890682,
90
- "f1": 0.7834666661687646,
112
+ "precision": 0.7446153846150028,
113
+ "recall": 0.8335246842704744,
114
+ "f1": 0.7865655466300883,
91
115
  "support": 1742
92
116
  }
93
117
  },
@@ -101,56 +125,56 @@
101
125
  {
102
126
  "low": 0.1,
103
127
  "high": 0.2,
104
- "n": 19,
105
- "acc": 0.5789473684210527
128
+ "n": 21,
129
+ "acc": 0.23809523809523808
106
130
  },
107
131
  {
108
132
  "low": 0.2,
109
133
  "high": 0.3,
110
- "n": 316,
111
- "acc": 0.3322784810126582
134
+ "n": 666,
135
+ "acc": 0.2912912912912913
112
136
  },
113
137
  {
114
138
  "low": 0.3,
115
139
  "high": 0.4,
116
- "n": 1213,
117
- "acc": 0.37757625721352017
140
+ "n": 2416,
141
+ "acc": 0.3265728476821192
118
142
  },
119
143
  {
120
144
  "low": 0.4,
121
145
  "high": 0.5,
122
- "n": 2754,
123
- "acc": 0.34132171387073346
146
+ "n": 4308,
147
+ "acc": 0.3152274837511606
124
148
  },
125
149
  {
126
150
  "low": 0.5,
127
151
  "high": 0.6,
128
- "n": 3647,
129
- "acc": 0.35481217438990953
152
+ "n": 4777,
153
+ "acc": 0.3382876282185472
130
154
  },
131
155
  {
132
156
  "low": 0.6,
133
157
  "high": 0.7,
134
- "n": 3659,
135
- "acc": 0.3883574747198688
158
+ "n": 4943,
159
+ "acc": 0.3568682986040866
136
160
  },
137
161
  {
138
162
  "low": 0.7,
139
163
  "high": 0.8,
140
- "n": 4610,
141
- "acc": 0.40629067245119305
164
+ "n": 5534,
165
+ "acc": 0.382544271774485
142
166
  },
143
167
  {
144
168
  "low": 0.8,
145
169
  "high": 0.9,
146
- "n": 7716,
147
- "acc": 0.43960601347848627
170
+ "n": 8066,
171
+ "acc": 0.43627572526655095
148
172
  },
149
173
  {
150
174
  "low": 0.9,
151
175
  "high": 1.0,
152
- "n": 37314,
153
- "acc": 0.5987564989012167
176
+ "n": 30517,
177
+ "acc": 0.6440344725890488
154
178
  }
155
179
  ]
156
180
  },
@@ -159,7 +183,7 @@
159
183
  "particle-honorific kryptonite (e.g. FR 'Saint-Just-Saint-Rambert') if not in synth set",
160
184
  "non-Latin scripts (CJK, Cyrillic) fall through to byte-fallback tokens; F1 unknown"
161
185
  ],
162
- "notes": "Stage 2 v3.0.0 \u2014 same encoder geometry as v2.0.x (8.87M params, 6L/256H/4-heads) plus a linear-chain CRF decoder (+~500 params with a frozen BIO transition mask), label smoothing on the per-token CE leg (disabled in the shipped hparams after iteration; see ship notes), and a 21-label classifier head (was 15) that adds venue / street / house_number BIO classes. Trained on corpus-v0.3.0 which adds the US DOT NAD source (~97M structured 911-grade address points). The CRF transition mask makes orphan-I sequences (e.g. \"Saint Petersburg \u2192 Petersburg\" clipping visible on the v0.2.0 demo) structurally impossible. See evals/scores-by-version.json for the v2.0.x \u2192 v3.0.0 deltas + the per-component F1 on the new fine labels.",
186
+ "notes": "v0.4.0 \u2014 issue #116. Same encoder geometry as v0.3.0 (8.87M params, 6L/256H/4-heads, 21 BIO labels, linear-chain CRF). Issue proposed (1) per-token CRF NLL normalization + (3) class-weighted CE biased toward coarse labels + (4) source-weight rebalance. Empirical iteration found that \u00a71 and \u00a73 destabilize sustained training at every LR tested (5e-4, 3e-4, 1.5e-4 \u2014 even the v0.3.0-safe LR), and on the golden v0.1.2 eval (4535 entries) they slightly REGRESS country/postcode F1 vs v0.3.0. SHIPPED recipe is the \u00a74-only ablation (v0.3.0 dual-loss + v0.4.0 source-weight rebalance) at lr=1.5e-4, step 2200. Modest fine-label gains (street +0.03, house_number +0.01), modest coarse-F1 regression (country -0.07, postcode -0.07). \u00a71/\u00a73 deferred to v0.4.1 corpus-side investigation per issue's '2K divergence' clause. Full ablation matrix retrospective in LOG.md.",
163
187
  "format": {
164
188
  "model": "ONNX int8 dynamic",
165
189
  "tokenizer": "SentencePiece unigram, byte_fallback=true, vocab_size=16000",
@@ -171,5 +195,5 @@
171
195
  "tokenizer": "tokenizer.model",
172
196
  "model_card": "model-card.json"
173
197
  },
174
- "base_relpath": "/data/models/checkpoints/stage2/step-001800"
198
+ "base_relpath": "/data/models/checkpoints/v0_4_0-stableLR-source-only/step-002200"
175
199
  }
package/model.onnx CHANGED
Binary file
package/package.json CHANGED
@@ -1,20 +1,19 @@
1
1
  {
2
- "name": "@mailwoman/neural-weights-fr-fr",
3
- "version": "3.0.0",
4
- "description": "Mailwoman neural-classifier weights for locale 'fr-fr'. Data-only package — loaded by @mailwoman/neural at runtime.",
5
- "license": "AGPL-3.0-only",
6
- "private": false,
7
- "repository": {
8
- "type": "git",
9
- "url": "https://github.com/sister-software/mailwoman"
10
- },
11
- "files": [
12
- "model.onnx",
13
- "tokenizer.model",
14
- "model-card.json",
15
- "README.md"
16
- ],
17
- "publishConfig": {
18
- "access": "public"
19
- }
20
- }
2
+ "name": "@mailwoman/neural-weights-fr-fr",
3
+ "version": "4.0.0",
4
+ "description": "Mailwoman neural-classifier weights for locale 'fr-fr'. Data-only package — loaded by @mailwoman/neural at runtime.",
5
+ "license": "AGPL-3.0-only",
6
+ "repository": {
7
+ "type": "git",
8
+ "url": "https://github.com/sister-software/mailwoman"
9
+ },
10
+ "files": [
11
+ "model.onnx",
12
+ "tokenizer.model",
13
+ "model-card.json",
14
+ "README.md"
15
+ ],
16
+ "publishConfig": {
17
+ "access": "public"
18
+ }
19
+ }
package/tokenizer.model CHANGED
Binary file