geberta-base-finetuned-augmentation

This model is a fine-tuned version of ikim-uk-essen/geberta-base on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.4327
• F1: 0.6097
• Roc Auc: 0.7506
• Accuracy: 0.4563

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
0.3808	1.0	141	0.3997	0.2027	0.5729	0.3119
0.3278	2.0	282	0.3627	0.3647	0.6353	0.3939
0.2881	3.0	423	0.3447	0.4099	0.6583	0.4349
0.2479	4.0	564	0.3317	0.4440	0.6741	0.4456
0.1888	5.0	705	0.3475	0.5081	0.6974	0.4439
0.135	6.0	846	0.3659	0.5597	0.7345	0.4332
0.1031	7.0	987	0.3894	0.5817	0.7401	0.4635
0.0755	8.0	1128	0.4100	0.5799	0.7292	0.4510
0.0559	9.0	1269	0.4327	0.6097	0.7506	0.4563
0.041	10.0	1410	0.4568	0.5988	0.7464	0.4456
0.0247	11.0	1551	0.4807	0.5891	0.7399	0.4456
0.0188	12.0	1692	0.5030	0.5945	0.7443	0.4403
0.0169	13.0	1833	0.5272	0.6055	0.7508	0.4510

Framework versions

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