|
--- |
|
license: apache-2.0 |
|
base_model: distilbert-base-uncased |
|
tags: |
|
- generated_from_keras_callback |
|
model-index: |
|
- name: NourhanAbosaeed/Coursera_Reviews_Sentiment_Analysis_DistillBERT |
|
results: [] |
|
language: |
|
- en |
|
library_name: transformers |
|
--- |
|
|
|
<!-- This model card has been generated automatically according to the information Keras had access to. You should |
|
probably proofread and complete it, then remove this comment. --> |
|
|
|
# NourhanAbosaeed/Coursera_Reviews_Sentiment_Analysis_DistillBERT |
|
|
|
This model is a fine-tuned version of [distilbert-base-uncased](https://huggingface.co/distilbert-base-uncased) on an a dataset from Cousera courses reviews, |
|
It is publicly available on Kaggle since 2017. |
|
|
|
After data preprocessing and model training, It achieves the following results on the evaluation set: |
|
- Train Loss: 0.4934 |
|
- Validation Loss: 0.6018 |
|
- Train Accuracy: 0.7498 |
|
- Epoch: 2 |
|
|
|
|
|
Considering the imbalanced nature of the data, metrics such as recall, precision, and F1 score were employed for evaluation:- |
|
|
|
The model achieves these results on the test set: |
|
|
|
precision recall f1-score support |
|
|
|
0 0.37 0.59 0.46 1928 |
|
1 0.71 0.74 0.72 1022 |
|
2 0.91 0.79 0.85 8712 |
|
|
|
accuracy 0.75 11662 |
|
macro avg 0.67 0.71 0.68 11662 |
|
weighted avg 0.81 0.75 0.77 11662 |
|
|
|
|
|
|
|
## Model description |
|
|
|
More information needed |
|
|
|
## Intended uses & limitations |
|
|
|
More information needed |
|
|
|
## Training and evaluation data |
|
|
|
After cleaning the data, It becomes 93291 training size, 11661 for validation and 11662 for test sets. |
|
|
|
#### There are 3 lables positive, negative and netural. |
|
|
|
### The data have imbalanced nature so I have used class weights during training. |
|
|
|
## Training procedure |
|
|
|
### Training hyperparameters |
|
|
|
The following hyperparameters were used during training: |
|
- optimizer: {'name': 'Adam', 'weight_decay': None, 'clipnorm': None, 'global_clipnorm': None, 'clipvalue': None, 'use_ema': False, 'ema_momentum': 0.99, 'ema_overwrite_frequency': None, 'jit_compile': True, 'is_legacy_optimizer': False, 'learning_rate': {'module': 'keras.optimizers.schedules', 'class_name': 'PolynomialDecay', 'config': {'initial_learning_rate': 2e-05, 'decay_steps': 17490, 'end_learning_rate': 0.0, 'power': 1.0, 'cycle': False, 'name': None}, 'registered_name': None}, 'beta_1': 0.9, 'beta_2': 0.999, 'epsilon': 1e-08, 'amsgrad': False} |
|
- training_precision: float32 |
|
|
|
### Training results |
|
|
|
| Train Loss | Validation Loss | Train Accuracy | Epoch | |
|
|:----------:|:---------------:|:--------------:|:-----:| |
|
| 0.6870 | 0.6382 | 0.7505 | 0 | |
|
| 0.5836 | 0.5976 | 0.7583 | 1 | |
|
| 0.4934 | 0.6018 | 0.7498 | 2 | |
|
|
|
|
|
### Framework versions |
|
|
|
- Transformers 4.35.1 |
|
- TensorFlow 2.14.0 |
|
- Datasets 2.14.7 |
|
- Tokenizers 0.14.1 |
