roberta-news-classifier

This model is a fine-tuned version of russellc/roberta-news-classifier on the custom(Kaggle) dataset. It achieves the following results on the evaluation set:

• Loss: 0.1043
• Accuracy: 0.9786
• F1: 0.9786
• Precision: 0.9786
• Recall: 0.9786

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: 1e-05
• train_batch_size: 32
• eval_batch_size: 64
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• num_epochs: 5

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1	Precision	Recall
0.1327	1.0	123	0.1043	0.9786	0.9786	0.9786	0.9786
0.1103	2.0	246	0.1157	0.9735	0.9735	0.9735	0.9735
0.102	3.0	369	0.1104	0.9735	0.9735	0.9735	0.9735
0.0825	4.0	492	0.1271	0.9714	0.9714	0.9714	0.9714
0.055	5.0	615	0.1296	0.9724	0.9724	0.9724	0.9724

Evaluation results

***** Running Prediction *****
Num examples = 980
Batch size = 64

          precision    recall  f1-score   support

  dunya        0.99      0.96      0.97       147  
ekonomi        0.96      0.96      0.96       141  
 kultur        0.97      0.99      0.98       142  
 saglik        0.99      0.98      0.98       148  
 siyaset       0.98      0.98      0.98       134  
 spor          1.00      1.00      1.00       139  
 teknoloji     0.96      0.98      0.97       129  
 accuracy      --        --        0.98       980
 macro avg     0.98      0.98      0.98       980  
 weighted avg  0.98      0.98      0.98       980 

   

Framework versions

• Transformers 4.25.1
• Pytorch 1.12.1+cu113
• Datasets 2.7.1
• Tokenizers 0.13.2