valueeval24-modern-bert-cos-initialfreeze-diff-lr-2

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

• Loss: 0.7800
• F1: 0.2007
• Roc Auc: 0.5700
• Accuracy: 0.1008

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: 3e-05
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.5
• num_epochs: 20
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	F1	Roc Auc	Accuracy
0.0167	1.0	3115	0.1945	0.1955	0.5655	0.1004
0.0146	2.0	6230	0.2180	0.1987	0.5679	0.1013
0.0133	3.0	9345	0.2485	0.2006	0.5698	0.0991
0.0114	4.0	12460	0.2969	0.1893	0.5627	0.0950
0.0101	5.0	15575	0.3334	0.1969	0.5684	0.0973
0.0082	6.0	18690	0.4324	0.1971	0.5682	0.0995
0.0065	7.0	21805	0.6038	0.1950	0.5677	0.0959
0.0043	8.0	24920	0.6648	0.2028	0.5731	0.0978
0.0033	9.0	28035	0.6840	0.2052	0.5725	0.1030
0.0015	10.0	31150	0.7048	0.2037	0.5728	0.1016
0.0009	11.0	34265	0.7291	0.1944	0.5667	0.0987
0.0006	12.0	37380	0.7446	0.1990	0.5693	0.0993
0.0004	13.0	40495	0.7531	0.2017	0.5712	0.1004
0.0003	14.0	43610	0.7655	0.1987	0.5684	0.1008
0.0002	15.0	46725	0.7750	0.1976	0.5680	0.0988
0.0002	16.0	49840	0.7800	0.2007	0.5700	0.1008

Framework versions

• Transformers 4.53.1
• Pytorch 2.6.0+cu124
• Datasets 2.14.4
• Tokenizers 0.21.2