model_from_berturk_1401_v3

This model is a fine-tuned version of dbmdz/bert-base-turkish-cased on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.4042
• Precision: 0.8896
• Recall: 0.8841
• F1: 0.8868
• Accuracy: 0.9198

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: 32
• eval_batch_size: 32
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 15

Training results

Training Loss	Epoch	Step	Validation Loss	Precision	Recall	F1	Accuracy
No log	1.0	244	0.3948	0.8565	0.8508	0.8536	0.8932
No log	2.0	488	0.3331	0.8724	0.8663	0.8693	0.9060
0.6019	3.0	732	0.3061	0.8855	0.8746	0.8800	0.9147
0.6019	4.0	976	0.3025	0.8881	0.8828	0.8855	0.9177
0.2753	5.0	1220	0.3137	0.8807	0.8819	0.8813	0.9148
0.2753	6.0	1464	0.3140	0.8876	0.8854	0.8865	0.9178
0.1963	7.0	1708	0.3210	0.8871	0.8840	0.8855	0.9182
0.1963	8.0	1952	0.3304	0.8908	0.8855	0.8882	0.9208
0.1431	9.0	2196	0.3452	0.8907	0.8843	0.8875	0.9206
0.1431	10.0	2440	0.3584	0.8896	0.8835	0.8865	0.9201
0.1061	11.0	2684	0.3770	0.8883	0.8849	0.8866	0.9191
0.1061	12.0	2928	0.3852	0.8876	0.8834	0.8855	0.9186
0.082	13.0	3172	0.3941	0.8894	0.8833	0.8863	0.9195
0.082	14.0	3416	0.3973	0.8893	0.8842	0.8867	0.9197
0.0694	15.0	3660	0.4042	0.8896	0.8841	0.8868	0.9198

Framework versions

• Transformers 4.25.1
• Pytorch 1.13.0+cu116
• Datasets 2.8.0
• Tokenizers 0.13.2