my_distilbert_model

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

• Define and compile the BERT and DistilBERT models for sentiment classification.
• Configure the model's different hyperparameters (learning rate, batch size, number of epochs) to find an optimal configuration.
• TensorFlow Datasets: Training and testing data were converted into TensorFlow datasets for efficient training.
• Early Stopping: Early stopping was implemented to prevent overfitting by monitoring validation loss.
• Compile the model
• The models were trained on the prepared training dataset in batches.

Training hyperparameters

• 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': 2e-05, 'beta_1': 0.9, 'beta_2': 0.999, 'epsilon': 1e-08, 'amsgrad': False}
• training_precision: float32

Training results

• Accuracy: 91% of the reviews in the test set were classified correctly. This is similar to the accuracy achieved by the BERT model (91%).
• Precision:
  • Negative Class: 89% of the predicted negative reviews were actually negative.
  • Positive Class: 93% of the predicted positive reviews were actually positive.
• Recall:
  • Negative Class: 93% of the actual negative reviews were correctly classified.
  • Positive Class: 89% of the actual positive reviews were correctly classified.
• F1-Score: The F1-score for both classes is around 0.91, indicating a good balance between precision and recall.

Framework versions

• Transformers 4.38.2
• TensorFlow 2.15.1
• Tokenizers 0.15.2