Auto-improve cycle 8 — F1=95.7%
Browse files- README.md +26 -81
- checkpoint-10116/config.json +47 -0
- checkpoint-10116/model.safetensors +3 -0
- checkpoint-10116/optimizer.pt +3 -0
- checkpoint-10116/rng_state.pth +3 -0
- checkpoint-10116/scheduler.pt +3 -0
- checkpoint-10116/trainer_state.json +0 -0
- checkpoint-10116/training_args.bin +3 -0
- checkpoint-8430/config.json +47 -0
- checkpoint-8430/model.safetensors +3 -0
- checkpoint-8430/optimizer.pt +3 -0
- checkpoint-8430/rng_state.pth +3 -0
- checkpoint-8430/scheduler.pt +3 -0
- checkpoint-8430/trainer_state.json +0 -0
- checkpoint-8430/training_args.bin +3 -0
- model.safetensors +1 -1
README.md
CHANGED
|
@@ -43,19 +43,9 @@ model-index:
|
|
| 43 |
- name: F1 Micro
|
| 44 |
type: f1
|
| 45 |
value: 0.970
|
| 46 |
-
verified: false
|
| 47 |
- name: F1 Macro
|
| 48 |
type: f1
|
| 49 |
-
value: 0.
|
| 50 |
-
verified: false
|
| 51 |
-
- name: Precision
|
| 52 |
-
type: precision
|
| 53 |
-
value: 0.953
|
| 54 |
-
verified: false
|
| 55 |
-
- name: Recall
|
| 56 |
-
type: recall
|
| 57 |
-
value: 0.989
|
| 58 |
-
verified: false
|
| 59 |
---
|
| 60 |
|
| 61 |
# Egide Toxicity Model
|
|
@@ -64,7 +54,7 @@ Modele de detection de toxicite multilingue (francais/anglais) concu pour la mod
|
|
| 64 |
|
| 65 |
## Description
|
| 66 |
|
| 67 |
-
Ce modele a ete
|
| 68 |
|
| 69 |
Le modele detecte **6 categories de toxicite** sans aucune regle codee en dur : tout repose sur l'inference IA.
|
| 70 |
|
|
@@ -81,51 +71,26 @@ Le modele detecte **6 categories de toxicite** sans aucune regle codee en dur :
|
|
| 81 |
|
| 82 |
## Performance
|
| 83 |
|
| 84 |
-
|
| 85 |
-
|
| 86 |
-
| Metrique | Score |
|
| 87 |
-
|---|---|
|
| 88 |
-
| **F1 Micro** | **97.0%** |
|
| 89 |
-
| **F1 Macro** | **96.6%** |
|
| 90 |
-
| **Precision** | **95.3%** |
|
| 91 |
-
| **Recall** | **98.9%** |
|
| 92 |
-
|
| 93 |
-
### F1 par categorie
|
| 94 |
|
| 95 |
| Categorie | F1 Score |
|
| 96 |
|---|---|
|
| 97 |
-
| toxicity |
|
| 98 |
-
| insult |
|
| 99 |
-
| hate |
|
| 100 |
-
| sexual |
|
| 101 |
-
| threat |
|
| 102 |
-
| identity_attack |
|
| 103 |
-
|
| 104 |
-
|
| 105 |
-
|
| 106 |
-
Teste avec un benchmark de 1000 messages simulant un chat Twitch reel (messages propres, insultes, haine, sexisme, menaces, doxxing, homophobie, evasion leet-speak, pieges a faux positifs) :
|
| 107 |
-
|
| 108 |
-
| Metrique | Score |
|
| 109 |
-
|---|---|
|
| 110 |
-
| **Accuracy** | **95.8%** |
|
| 111 |
-
| **Precision** | **95.9%** |
|
| 112 |
-
| **Recall** | **96.3%** |
|
| 113 |
-
| **F1 Score** | **96.1%** |
|
| 114 |
-
| True Positives | 520 |
|
| 115 |
-
| True Negatives | 438 |
|
| 116 |
-
| False Positives | 22 |
|
| 117 |
-
| False Negatives | 20 |
|
| 118 |
-
|
| 119 |
-
**Categories a 100% de precision** : greetings, compliments, questions, reactions, casual, doxxing, hate, homophobie, menaces, toxicite subtile, pieges a faux positifs.
|
| 120 |
|
| 121 |
## Points forts
|
| 122 |
|
| 123 |
- **Multilingue** : Comprend le francais et l'anglais nativement
|
| 124 |
- **Texte obfusque** : Detecte les insultes deguisees comme "ntm", "n t m", "fdp", "f.d.p", "c0nn4rd", "k y s", etc.
|
| 125 |
- **Contexte gaming** : Ne flag PAS les expressions figuratives courantes en gaming ("ca tue ce jeu", "je suis mort de rire", "this game is killing me")
|
| 126 |
-
- **Argot Twitch** : Entraine sur du
|
| 127 |
- **Zero faux pattern** : Aucune regex, aucune liste de mots interdits, 100% IA
|
| 128 |
-
- **Perte ponderee** : Utilise BCEWithLogitsLoss avec pos_weight pour gerer le desequilibre de classes
|
| 129 |
|
| 130 |
## Utilisation
|
| 131 |
|
|
@@ -177,40 +142,20 @@ curl -X POST http://localhost:8000/analyze \
|
|
| 177 |
|
| 178 |
## Entrainement
|
| 179 |
|
| 180 |
-
- **Modele de base** : `xlm-roberta-base` (278M parametres)
|
| 181 |
- **Type** : Multi-label classification
|
| 182 |
-
- **Loss** : BCEWithLogitsLoss
|
| 183 |
-
- **
|
| 184 |
-
- **Epochs** : 12
|
| 185 |
- **Batch size** : 8
|
| 186 |
- **Learning rate** : 2e-5
|
| 187 |
- **Warmup** : 10%
|
| 188 |
-
- **
|
| 189 |
-
-
|
| 190 |
-
-
|
| 191 |
-
|
| 192 |
-
|
| 193 |
-
|
| 194 |
-
|
| 195 |
-
|
| 196 |
-
- Insultes anglaises (standard + obfusquees)
|
| 197 |
-
- Doxxing / partage d'infos personnelles
|
| 198 |
-
- Toxicite subtile / passive-aggressive
|
| 199 |
-
- **4285 exemples non-toxiques (87.5%)** :
|
| 200 |
-
- 3638 messages reels de chat Twitch collectes sur 56 chaines live (faux positifs confirmes + messages propres)
|
| 201 |
-
- 382 exemples classiques non-toxiques
|
| 202 |
-
- 265 exemples d'argot Twitch (emotes, abreviations, greetings, slang)
|
| 203 |
-
|
| 204 |
-
### Poids positifs par label (gestion du desequilibre)
|
| 205 |
-
|
| 206 |
-
| Label | Poids | Positifs | Negatifs |
|
| 207 |
-
|---|---|---|---|
|
| 208 |
-
| toxicity | 6.98 | 1565 | 10922 |
|
| 209 |
-
| insult | 10.00 | 876 | 11611 |
|
| 210 |
-
| hate | 10.00 | 456 | 12031 |
|
| 211 |
-
| sexual | 10.00 | 95 | 12392 |
|
| 212 |
-
| threat | 10.00 | 192 | 12295 |
|
| 213 |
-
| identity_attack | 10.00 | 332 | 12155 |
|
| 214 |
|
| 215 |
## Architecture du projet Egide
|
| 216 |
|
|
@@ -222,9 +167,9 @@ Le bot Node.js envoie les messages au service Python via HTTP. Le service charge
|
|
| 222 |
|
| 223 |
## Limitations
|
| 224 |
|
| 225 |
-
- Entraine principalement sur du francais et de l'anglais. D'autres langues peuvent fonctionner
|
| 226 |
-
-
|
| 227 |
-
- Les
|
| 228 |
|
| 229 |
## Licence
|
| 230 |
|
|
@@ -236,7 +181,7 @@ Apache 2.0
|
|
| 236 |
@misc{egide-toxicity-model,
|
| 237 |
author = {Loule},
|
| 238 |
title = {Egide Toxicity Model - Multilingual Toxicity Detection for Twitch Chat},
|
| 239 |
-
year = {
|
| 240 |
publisher = {HuggingFace},
|
| 241 |
url = {https://huggingface.co/Loule/egide-toxicity-model}
|
| 242 |
}
|
|
|
|
| 43 |
- name: F1 Micro
|
| 44 |
type: f1
|
| 45 |
value: 0.970
|
|
|
|
| 46 |
- name: F1 Macro
|
| 47 |
type: f1
|
| 48 |
+
value: 0.969
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 49 |
---
|
| 50 |
|
| 51 |
# Egide Toxicity Model
|
|
|
|
| 54 |
|
| 55 |
## Description
|
| 56 |
|
| 57 |
+
Ce modele a ete entraine pour la classification multi-label de contenu toxique. Il a ete concu specifiquement pour le projet [Egide](https://github.com/Loule95450/Egide), un bot de moderation Twitch alimente par l'IA.
|
| 58 |
|
| 59 |
Le modele detecte **6 categories de toxicite** sans aucune regle codee en dur : tout repose sur l'inference IA.
|
| 60 |
|
|
|
|
| 71 |
|
| 72 |
## Performance
|
| 73 |
|
| 74 |
+
Evalue sur un jeu de test de 243 exemples (15% du dataset) :
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 75 |
|
| 76 |
| Categorie | F1 Score |
|
| 77 |
|---|---|
|
| 78 |
+
| **toxicity** | 0.981 |
|
| 79 |
+
| **insult** | 0.974 |
|
| 80 |
+
| **hate** | 0.949 |
|
| 81 |
+
| **sexual** | 1.000 |
|
| 82 |
+
| **threat** | 0.966 |
|
| 83 |
+
| **identity_attack** | 0.945 |
|
| 84 |
+
| **F1 Micro** | **0.970** |
|
| 85 |
+
| **F1 Macro** | **0.969** |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 86 |
|
| 87 |
## Points forts
|
| 88 |
|
| 89 |
- **Multilingue** : Comprend le francais et l'anglais nativement
|
| 90 |
- **Texte obfusque** : Detecte les insultes deguisees comme "ntm", "n t m", "fdp", "f.d.p", "c0nn4rd", "k y s", etc.
|
| 91 |
- **Contexte gaming** : Ne flag PAS les expressions figuratives courantes en gaming ("ca tue ce jeu", "je suis mort de rire", "this game is killing me")
|
| 92 |
+
- **Argot Twitch** : Entraine sur du vocabulaire de chat Twitch (emotes, abreviations, slang)
|
| 93 |
- **Zero faux pattern** : Aucune regex, aucune liste de mots interdits, 100% IA
|
|
|
|
| 94 |
|
| 95 |
## Utilisation
|
| 96 |
|
|
|
|
| 142 |
|
| 143 |
## Entrainement
|
| 144 |
|
|
|
|
| 145 |
- **Type** : Multi-label classification
|
| 146 |
+
- **Loss** : BCEWithLogitsLoss
|
| 147 |
+
- **Epochs** : 10 (best model a l'epoch 7)
|
|
|
|
| 148 |
- **Batch size** : 8
|
| 149 |
- **Learning rate** : 2e-5
|
| 150 |
- **Warmup** : 10%
|
| 151 |
+
- **Dataset** : 539 exemples curates x3 augmentation = 1617 exemples
|
| 152 |
+
- Insultes francaises (standard + obfusquees)
|
| 153 |
+
- Discours de haine (racisme, xenophobie, antisemitisme)
|
| 154 |
+
- Sexisme
|
| 155 |
+
- Homophobie / transphobie
|
| 156 |
+
- Menaces (standard + obfusquees)
|
| 157 |
+
- Insultes anglaises (standard + obfusquees)
|
| 158 |
+
- ~150 exemples non-toxiques (chat Twitch, expressions figuratives, emotes)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 159 |
|
| 160 |
## Architecture du projet Egide
|
| 161 |
|
|
|
|
| 167 |
|
| 168 |
## Limitations
|
| 169 |
|
| 170 |
+
- Entraine principalement sur du francais et de l'anglais. D'autres langues peuvent fonctionner mais avec moins de precision.
|
| 171 |
+
- Le dataset d'entrainement est relativement petit (539 exemples uniques). Des ameliorations sont possibles en ajoutant plus de donnees.
|
| 172 |
+
- Les nouvelles formes d'obfuscation non vues a l'entrainement peuvent echapper a la detection.
|
| 173 |
|
| 174 |
## Licence
|
| 175 |
|
|
|
|
| 181 |
@misc{egide-toxicity-model,
|
| 182 |
author = {Loule},
|
| 183 |
title = {Egide Toxicity Model - Multilingual Toxicity Detection for Twitch Chat},
|
| 184 |
+
year = {2025},
|
| 185 |
publisher = {HuggingFace},
|
| 186 |
url = {https://huggingface.co/Loule/egide-toxicity-model}
|
| 187 |
}
|
checkpoint-10116/config.json
ADDED
|
@@ -0,0 +1,47 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"add_cross_attention": false,
|
| 3 |
+
"architectures": [
|
| 4 |
+
"XLMRobertaForSequenceClassification"
|
| 5 |
+
],
|
| 6 |
+
"attention_probs_dropout_prob": 0.1,
|
| 7 |
+
"bos_token_id": 0,
|
| 8 |
+
"classifier_dropout": null,
|
| 9 |
+
"dtype": "float32",
|
| 10 |
+
"eos_token_id": 2,
|
| 11 |
+
"hidden_act": "gelu",
|
| 12 |
+
"hidden_dropout_prob": 0.1,
|
| 13 |
+
"hidden_size": 768,
|
| 14 |
+
"id2label": {
|
| 15 |
+
"0": "LABEL_0",
|
| 16 |
+
"1": "LABEL_1",
|
| 17 |
+
"2": "LABEL_2",
|
| 18 |
+
"3": "LABEL_3",
|
| 19 |
+
"4": "LABEL_4",
|
| 20 |
+
"5": "LABEL_5"
|
| 21 |
+
},
|
| 22 |
+
"initializer_range": 0.02,
|
| 23 |
+
"intermediate_size": 3072,
|
| 24 |
+
"is_decoder": false,
|
| 25 |
+
"label2id": {
|
| 26 |
+
"LABEL_0": 0,
|
| 27 |
+
"LABEL_1": 1,
|
| 28 |
+
"LABEL_2": 2,
|
| 29 |
+
"LABEL_3": 3,
|
| 30 |
+
"LABEL_4": 4,
|
| 31 |
+
"LABEL_5": 5
|
| 32 |
+
},
|
| 33 |
+
"layer_norm_eps": 1e-05,
|
| 34 |
+
"max_position_embeddings": 514,
|
| 35 |
+
"model_type": "xlm-roberta",
|
| 36 |
+
"num_attention_heads": 12,
|
| 37 |
+
"num_hidden_layers": 12,
|
| 38 |
+
"output_past": true,
|
| 39 |
+
"pad_token_id": 1,
|
| 40 |
+
"position_embedding_type": "absolute",
|
| 41 |
+
"problem_type": "multi_label_classification",
|
| 42 |
+
"tie_word_embeddings": true,
|
| 43 |
+
"transformers_version": "5.2.0",
|
| 44 |
+
"type_vocab_size": 1,
|
| 45 |
+
"use_cache": false,
|
| 46 |
+
"vocab_size": 250002
|
| 47 |
+
}
|
checkpoint-10116/model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ce07fe7999dc347dff7dbefe7220f22e45dd7bbfd84f014df51343f9841025a3
|
| 3 |
+
size 1112217288
|
checkpoint-10116/optimizer.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8860905400159773bd8e306232ce3d7286bec2fef74c2e47659b4362133745f5
|
| 3 |
+
size 2224549003
|
checkpoint-10116/rng_state.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e38e4bded6fa666a78b369223010c84f8eafdd4ce4069224aa6f2854b4222440
|
| 3 |
+
size 14455
|
checkpoint-10116/scheduler.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d387f002bc7eb25457dd64c63b2422eb747ab631ed7ebc80279b42718ab00662
|
| 3 |
+
size 1465
|
checkpoint-10116/trainer_state.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
checkpoint-10116/training_args.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ebdc3d07a5dca3e7b0152072589f926cb3878579b56582b670d1880fb774230e
|
| 3 |
+
size 5265
|
checkpoint-8430/config.json
ADDED
|
@@ -0,0 +1,47 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"add_cross_attention": false,
|
| 3 |
+
"architectures": [
|
| 4 |
+
"XLMRobertaForSequenceClassification"
|
| 5 |
+
],
|
| 6 |
+
"attention_probs_dropout_prob": 0.1,
|
| 7 |
+
"bos_token_id": 0,
|
| 8 |
+
"classifier_dropout": null,
|
| 9 |
+
"dtype": "float32",
|
| 10 |
+
"eos_token_id": 2,
|
| 11 |
+
"hidden_act": "gelu",
|
| 12 |
+
"hidden_dropout_prob": 0.1,
|
| 13 |
+
"hidden_size": 768,
|
| 14 |
+
"id2label": {
|
| 15 |
+
"0": "LABEL_0",
|
| 16 |
+
"1": "LABEL_1",
|
| 17 |
+
"2": "LABEL_2",
|
| 18 |
+
"3": "LABEL_3",
|
| 19 |
+
"4": "LABEL_4",
|
| 20 |
+
"5": "LABEL_5"
|
| 21 |
+
},
|
| 22 |
+
"initializer_range": 0.02,
|
| 23 |
+
"intermediate_size": 3072,
|
| 24 |
+
"is_decoder": false,
|
| 25 |
+
"label2id": {
|
| 26 |
+
"LABEL_0": 0,
|
| 27 |
+
"LABEL_1": 1,
|
| 28 |
+
"LABEL_2": 2,
|
| 29 |
+
"LABEL_3": 3,
|
| 30 |
+
"LABEL_4": 4,
|
| 31 |
+
"LABEL_5": 5
|
| 32 |
+
},
|
| 33 |
+
"layer_norm_eps": 1e-05,
|
| 34 |
+
"max_position_embeddings": 514,
|
| 35 |
+
"model_type": "xlm-roberta",
|
| 36 |
+
"num_attention_heads": 12,
|
| 37 |
+
"num_hidden_layers": 12,
|
| 38 |
+
"output_past": true,
|
| 39 |
+
"pad_token_id": 1,
|
| 40 |
+
"position_embedding_type": "absolute",
|
| 41 |
+
"problem_type": "multi_label_classification",
|
| 42 |
+
"tie_word_embeddings": true,
|
| 43 |
+
"transformers_version": "5.2.0",
|
| 44 |
+
"type_vocab_size": 1,
|
| 45 |
+
"use_cache": false,
|
| 46 |
+
"vocab_size": 250002
|
| 47 |
+
}
|
checkpoint-8430/model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:567f20e8a633069d2f3739140dbd92d459f54141439aa81ec0dccb37cbd0169e
|
| 3 |
+
size 1112217288
|
checkpoint-8430/optimizer.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1d3fd5fe2d90bdd7ef2971bdc1ede0e8d98017088bd7b3e70ab8c556554108a1
|
| 3 |
+
size 2224549003
|
checkpoint-8430/rng_state.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2b652c6269b998b96ab924b2734c0818fab436c642524e13fc6cd4d9082e62b5
|
| 3 |
+
size 14455
|
checkpoint-8430/scheduler.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d983e9e82ac87acb66331cc9b5cfaa6ff89e902db822c7a81f3220ba99096811
|
| 3 |
+
size 1465
|
checkpoint-8430/trainer_state.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
checkpoint-8430/training_args.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ebdc3d07a5dca3e7b0152072589f926cb3878579b56582b670d1880fb774230e
|
| 3 |
+
size 5265
|
model.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 1112217288
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ce07fe7999dc347dff7dbefe7220f22e45dd7bbfd84f014df51343f9841025a3
|
| 3 |
size 1112217288
|