bert-base-uncased-uHack_reviews_multilabel_clf
This model is a fine-tuned version of bert-base-uncased. It achieves the following results on the evaluation set:
- Loss: 0.1515
- F1: 0.8737
- Roc Auc: 0.9146
- Accuracy: 0.5967
Model description
This is a multilabel classification model of whether different aspects of a product are mentioned in reviews.
For more information on how it was created, check out the following link: https://github.com/DunnBC22/NLP_Projects/blob/main/Multilabel%20Classification/Review%20Sentiments/Sentiment_Analysis%20-%20Using%20BERT%20Instead%20of%20DistilBERT.ipynb
Intended uses & limitations
This model is intended to demonstrate my ability to solve a complex problem using technology.
Training and evaluation data
Dataset Source: https://www.kaggle.com/datasets/mohamedziauddin/mh-uhack-sentiments?select=train.csv
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: 10
Training results
Training Loss | Epoch | Step | Validation Loss | F1 | Roc Auc | Accuracy |
---|---|---|---|---|---|---|
0.3409 | 1.0 | 305 | 0.2453 | 0.7717 | 0.8284 | 0.3787 |
0.2049 | 2.0 | 610 | 0.1906 | 0.8389 | 0.8796 | 0.5066 |
0.1519 | 3.0 | 915 | 0.1707 | 0.8595 | 0.8992 | 0.5443 |
0.1176 | 4.0 | 1220 | 0.1575 | 0.8592 | 0.9012 | 0.5459 |
0.093 | 5.0 | 1525 | 0.1544 | 0.8647 | 0.9034 | 0.5656 |
0.0752 | 6.0 | 1830 | 0.1550 | 0.8630 | 0.9084 | 0.5656 |
0.0629 | 7.0 | 2135 | 0.1515 | 0.8737 | 0.9146 | 0.5967 |
0.0538 | 8.0 | 2440 | 0.1487 | 0.8717 | 0.9145 | 0.5852 |
0.0481 | 9.0 | 2745 | 0.1502 | 0.8714 | 0.9145 | 0.5836 |
0.0447 | 10.0 | 3050 | 0.1499 | 0.8715 | 0.9154 | 0.5852 |
Framework versions
- Transformers 4.27.1
- Pytorch 1.13.1+cu116
- Datasets 2.10.1
- Tokenizers 0.13.2