model_from_berturk_upos_22Jan_val_on_dev_v1

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: 0.4600
• Precision: 0.8894
• Recall: 0.8931
• F1: 0.8913
• Accuracy: 0.9217

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	170	0.3259	0.8864	0.8851	0.8857	0.9170
No log	2.0	340	0.2976	0.8991	0.9005	0.8998	0.9271
0.3859	3.0	510	0.2988	0.8970	0.8981	0.8976	0.9257
0.3859	4.0	680	0.3131	0.8977	0.8984	0.8981	0.9267
0.3859	5.0	850	0.3291	0.8943	0.8974	0.8958	0.9246
0.1338	6.0	1020	0.3298	0.8974	0.8984	0.8979	0.9253
0.1338	7.0	1190	0.3538	0.8908	0.8950	0.8929	0.9222
0.1338	8.0	1360	0.3879	0.8903	0.8921	0.8912	0.9220
0.0798	9.0	1530	0.3852	0.8929	0.8933	0.8931	0.9228
0.0798	10.0	1700	0.4111	0.8873	0.8896	0.8884	0.9196
0.0798	11.0	1870	0.4254	0.8928	0.8940	0.8934	0.9231
0.0507	12.0	2040	0.4476	0.8911	0.8930	0.8921	0.9224
0.0507	13.0	2210	0.4478	0.8900	0.8926	0.8913	0.9220
0.0507	14.0	2380	0.4563	0.8896	0.8931	0.8914	0.9217
0.0357	15.0	2550	0.4600	0.8894	0.8931	0.8913	0.9217

Framework versions

• Transformers 4.25.1
• Pytorch 1.13.1+cu116
• Datasets 2.8.0
• Tokenizers 0.13.2