text_classification_model
This model is a fine-tuned version of
distilbert/distilbert-base-uncased on an imdb dataset.
It achieves the following results on the evaluation set:
- Loss: 0.2291
- Accuracy: 0.9318
Model description
More information needed
Intended uses & limitations
this model is intended for binary sentiment classification on English-languange text especially in domains such as:
Movie Reviews (e.g IMDB),
Product or Service feedback,
Social media Posts and short user comments,
Customer Satisfaction surveys.
While the model performs well on movie review-style sentiment classification, it has the following limitations:
Language restriction: Trained only on English; not suitable for other languages
Domain Generalization: May perfom poorly on domains very different from movie reviews(e.g. medical, legal, or financial text)
Subtle Sentiment: May misclassify sarcasm, irony, or ambigous statements
Biases: inherits potential biases from IMDB dataset, including genre-related language patterns and reviewer tendencies
Input Length: Long texts are truncated to fit the model's maximun sequence length, which may impact classification accuracy on lengthy documents
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: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
- lr_scheduler_type: linear
- num_epochs: 2
Training results
| Training Loss | Epoch | Step | Validation Loss | Accuracy |
|---|
| 0.2235 | 1.0 | 1563 | 0.2121 | 0.9180 |
| 0.1481 | 2.0 | 3126 | 0.2291 | 0.9318 |
Framework versions
- Transformers 4.54.0
- Pytorch 2.6.0+cu124
- Datasets 4.0.0
- Tokenizers 0.21.2
