roberta-large-finetuned-augmentation-LUNAR-TAPT

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

• Loss: 0.4897
• F1: 0.8302
• Roc Auc: 0.8696
• Accuracy: 0.6338

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: cosine
• lr_scheduler_warmup_steps: 100
• num_epochs: 20

Training results

Training Loss	Epoch	Step	Validation Loss	F1	Roc Auc	Accuracy
0.3371	1.0	317	0.3025	0.7356	0.8000	0.5233
0.2571	2.0	634	0.3055	0.7376	0.7942	0.5572
0.1848	3.0	951	0.2850	0.7964	0.8431	0.5912
0.124	4.0	1268	0.3223	0.7738	0.8164	0.5635
0.0701	5.0	1585	0.3219	0.8091	0.8597	0.5951
0.0491	6.0	1902	0.3576	0.8148	0.8547	0.6014
0.0432	7.0	2219	0.3808	0.8216	0.8665	0.6196
0.0352	8.0	2536	0.3945	0.8278	0.8721	0.6259
0.0282	9.0	2853	0.4357	0.8173	0.8580	0.6054
0.012	10.0	3170	0.4670	0.8208	0.8679	0.5951
0.0054	11.0	3487	0.4864	0.8177	0.8599	0.6038
0.0029	12.0	3804	0.4882	0.8289	0.8687	0.6259
0.0011	13.0	4121	0.4897	0.8302	0.8696	0.6338
0.0012	14.0	4438	0.5079	0.8273	0.8680	0.6251
0.0008	15.0	4755	0.5146	0.8285	0.8688	0.6227
0.0007	16.0	5072	0.5100	0.8282	0.8693	0.6338
0.0008	17.0	5389	0.5158	0.8282	0.8673	0.6330

Framework versions

• Transformers 4.45.1
• Pytorch 2.4.0
• Datasets 3.0.1
• Tokenizers 0.20.0