<!-- 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. -->
bert-nlp-project-ft-news
This model is a fine-tuned version of jestemleon/bert-nlp-project-news on the news dataset. It achieves the following results on the evaluation set:
- Loss: 0.4362
- Accuracy: 0.9078
- F1: 0.8872
and flowing results on the testing set:
- Accuracy: 0.9001
- F1: 0.8827
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: 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: 3
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Accuracy | F1 |
|---|---|---|---|---|---|
| 0.4106 | 0.37 | 120 | 0.3049 | 0.8797 | 0.8444 |
| 0.3078 | 0.75 | 240 | 0.3038 | 0.8891 | 0.8683 |
| 0.2642 | 1.12 | 360 | 0.3363 | 0.8969 | 0.8769 |
| 0.2114 | 1.5 | 480 | 0.3460 | 0.8922 | 0.8671 |
| 0.1661 | 1.87 | 600 | 0.3894 | 0.9031 | 0.8826 |
| 0.156 | 2.24 | 720 | 0.3946 | 0.8953 | 0.8714 |
| 0.1079 | 2.62 | 840 | 0.4340 | 0.9016 | 0.8795 |
| 0.1083 | 2.99 | 960 | 0.4362 | 0.9078 | 0.8872 |
Framework versions
- Transformers 4.25.1
- Pytorch 1.13.0+cu116
- Datasets 2.8.0
- Tokenizers 0.13.2
