metadata
language: es
tags:
- generated_from_trainer
metrics:
- accuracy
model-index:
- name: roberta-base-bne-finetuned-ciberbullying-spanish
results:
- task:
name: Text Classification
type: text-classification
metrics:
- name: Accuracy
type: accuracy
value: 0.9607097303206997
roberta-base-bne-finetuned-ciberbullying-spanish
This model is a fine-tuned version of BSC-TeMU/roberta-base-bne on the dataset generated scrapping all social networks (Twitter, Youtube ...) to detect ciberbullying on Spanish.
It achieves the following results on the evaluation set:
- Loss: 0.1657
- Accuracy: 0.9607
Training and evaluation data
We use the concatenation from multiple datasets generated scrapping social networks (Twitter,Youtube,Discord...) to fine-tune this model. The total number of sentence pairs is above 360k sentences.
Training procedure
Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 2e-05
- train_batch_size: 16
- eval_batch_size: 16
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 4
Training results
| Training Loss | Epoch | Step | Accuracy | Validation Loss |
|---|---|---|---|---|
| 0.1512 | 1.0 | 22227 | 0.9501 | 0.1418 |
| 0.1253 | 2.0 | 44454 | 0.9567 | 0.1499 |
| 0.0973 | 3.0 | 66681 | 0.9594 | 0.1397 |
| 0.0658 | 4.0 | 88908 | 0.9607 | 0.1657 |
Framework versions
- Transformers 4.10.3
- Pytorch 1.9.0+cu102
- Datasets 1.12.1
- Tokenizers 0.10.3