DistilBERT base FR sexism detection
This model is a fine-tuned version of distilbert-base-multilingual-cased on the lidiapierre/fr_sexism_labelled dataset. It is intended to be used as a classification model for identifying sexist language in French (0 - not sexist; 1 - sexist).
It achieves the following results on the evaluation set:
- Loss: 0.3751
- Accuracy: 0.9123
- F1: 0.9206
Classification examples:
| Prediction | Text |
|---|---|
| sexist | Tu pourrais sourire plus |
| not sexist | Tout le monde à table |
Model description
Transformer-based language model for binary classification.
Risks & limitations
This model is susceptible of displaying bias inherited from its pretrained model: predictions generated by the model may include disturbing and harmful stereotypes across protected classes; identity characteristics; and sensitive, social, and occupational groups.
Training procedure
Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 5e-05
- train_batch_size: 8
- eval_batch_size: 8
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 3.0
Training results
| Epoch | Step | Validation Loss | Accuracy | F1 |
|---|---|---|---|---|
| 1.0 | 128 | 0.5027 | 0.8509 | 0.8759 |
| 2.0 | 256 | 0.2606 | 0.9298 | 0.9365 |
| 3.0 | 384 | 0.3751 | 0.9123 | 0.9206 |
Framework versions
- Transformers 4.34.0
- Pytorch 2.0.1+cu118
- Datasets 2.14.5
- Tokenizers 0.14.1
