<!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. -->
distilbert-base-uncased__hate_speech_offensive__train-16-9
This model is a fine-tuned version of distilbert-base-uncased on the None dataset. It achieves the following results on the evaluation set:
- Loss: 1.1121
- Accuracy: 0.16
Model description
More information needed
Intended uses & limitations
More information needed
Training and evaluation data
More information needed
Training procedure
Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 2e-05
- train_batch_size: 4
- eval_batch_size: 4
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 50
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Accuracy |
|---|---|---|---|---|
| 1.1038 | 1.0 | 10 | 1.1243 | 0.1 |
| 1.0859 | 2.0 | 20 | 1.1182 | 0.2 |
| 1.0234 | 3.0 | 30 | 1.1442 | 0.3 |
| 0.9493 | 4.0 | 40 | 1.2239 | 0.1 |
| 0.8114 | 5.0 | 50 | 1.2023 | 0.4 |
| 0.6464 | 6.0 | 60 | 1.2329 | 0.4 |
| 0.4731 | 7.0 | 70 | 1.2971 | 0.5 |
| 0.3355 | 8.0 | 80 | 1.3913 | 0.4 |
| 0.1268 | 9.0 | 90 | 1.4670 | 0.5 |
| 0.0747 | 10.0 | 100 | 1.7961 | 0.4 |
| 0.0449 | 11.0 | 110 | 1.8168 | 0.5 |
| 0.0307 | 12.0 | 120 | 1.9307 | 0.4 |
Framework versions
- Transformers 4.15.0
- Pytorch 1.10.2+cu102
- Datasets 1.18.2
- Tokenizers 0.10.3
