distilbert-bbc-news-classification

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: 0.0669
• Accuracy: 0.9880
• F1-score: 0.9880
• Recall: 0.9886
• Precision: 0.9875

Model description

More information needed

Intended uses & limitations

More information needed

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: 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-score	Recall	Precision
No log	1.0	98	0.1495	0.9775	0.9775	0.9772	0.9781
No log	2.0	196	0.0737	0.9880	0.9882	0.9885	0.9880
No log	3.0	294	0.0669	0.9880	0.9880	0.9886	0.9875

Framework versions

• Transformers 4.25.1
• Pytorch 1.13.1+cu116
• Datasets 2.8.0
• Tokenizers 0.13.2