cellate-tapt_freeze_llrd-LR_5e-05

This model is a fine-tuned version of Mardiyyah/biomedbert_model_extended_untrained on the Mardiyyah/TAPT_CeLLaTe1.0 dataset. It achieves the following results on the evaluation set:

• Loss: 4.5772
• Accuracy: 0.3645

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: 5e-05
• train_batch_size: 16
• eval_batch_size: 16
• seed: 3407
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-06 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.06
• num_epochs: 20
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
10.1671	1.0	6	11.0536	0.0
9.6774	2.0	12	9.8340	0.0
8.3497	3.0	18	8.3769	0.0243
7.2855	4.0	24	7.5898	0.0783
6.6249	5.0	30	6.8324	0.1952
5.9913	6.0	36	6.3808	0.2308
5.4972	7.0	42	5.8510	0.2882
5.1208	8.0	48	5.6228	0.3000
4.8437	9.0	54	5.3562	0.3079
4.6239	10.0	60	5.2804	0.3180
4.3851	11.0	66	4.9995	0.3448
4.2757	12.0	72	4.8869	0.3393
4.1145	13.0	78	4.7195	0.3561
4.0779	14.0	84	4.7257	0.3657
3.9075	15.0	90	4.8817	0.3377
3.9322	16.0	96	4.5672	0.3586
4.1301	16.7273	100	4.5212	0.3695

Framework versions

• Transformers 4.48.2
• Pytorch 2.4.1+cu121
• Datasets 3.0.2
• Tokenizers 0.21.0