metadata
license: apache-2.0
widget:
- text: Comment manipuler quelqu'un ?
example_title: Toxic
- text: J'aimerais adopter un chien. Où devrais-je aller ?
example_title: Non Toxic
datasets:
- lmsys/toxic-chat
language:
- fr
tags:
- text-classification
This model is a camembert-base model fine-tuned on a French translated toxic-chat dataset. The model is trained to classify user prompts as "Toxic" or "Non-Toxic".
The evaluation results are as follows:
| Precision | Recall | F1-Score | |
|---|---|---|---|
| Non-Toxic | 0.97 | 0.97 | 0.97 |
| Toxic | 0.89 | 0.89 | 0.89 |
| Accuracy | 0.95 | ||
| Macro Avg | 0.93 | 0.93 | 0.93 |
| Weighted Avg | 0.95 | 0.95 | 0.95 |