Roberta_coaid

This model is a fine-tuned version of adity12345/Roberta_covert on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.2850
• Accuracy: 0.897
• Auc: 0.862
• Precision: 0.905
• Recall: 0.984
• F1: 0.943
• F1-macro: 0.721
• F1-micro: 0.897
• F1-weighted: 0.881

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: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 32
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 2
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Auc	Precision	Recall	F1	F1-macro	F1-micro	F1-weighted
1.1024	1.0638	50	0.2850	0.897	0.862	0.905	0.984	0.943	0.721	0.897	0.881

Framework versions

• Transformers 4.55.2
• Pytorch 2.8.0+cu126
• Datasets 4.0.0
• Tokenizers 0.21.4