distilbert-base-uncased-SpamFilter-DunnBC22
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: 0.1007
- Accuracy: 0.9907
- F1: 0.9906
Model description
This is a binary classification of whether the inputs are spam or not.
For more information on how it was created, check out the following link: https://github.com/DunnBC22/NLP_Projects/blob/main/Binary%20Classification/Spam%20Filter%20-%20Smaller%20Dataset/DunnBC22-distilbert-base-uncased-SpamFilter-sm.ipynb
Intended uses & limitations
This model is intended to demonstrate my ability to solve a complex problem using technology.
The main limitation is the quality of the data source.
Training and evaluation data
Dataset Source: https://www.kaggle.com/datasets/datatattle/email-classification-nlp
Input Word Length By Class:
Confusion Matrix:
Training procedure
Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 2e-05
- train_batch_size: 128
- eval_batch_size: 128
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 3
Training results
| Training Loss | Epoch | Step | Validation Loss | Accuracy | F1 |
|---|---|---|---|---|---|
| 0.5039 | 1.0 | 7 | 0.3920 | 0.8333 | 0.7576 |
| 0.3008 | 2.0 | 14 | 0.2010 | 0.9722 | 0.9719 |
| 0.113 | 3.0 | 21 | 0.1007 | 0.9907 | 0.9906 |
Framework versions
- Transformers 4.21.1
- Pytorch 1.12.1
- Datasets 2.4.0
- Tokenizers 0.12.1
