berturk-ner

This model is a fine-tuned version of dbmdz/bert-base-turkish-cased on turkish-nlp-suite/turkish-wikiNER dataset. It achieves the following results:

Validation Set

• Loss: 0.3693
• Accuracy: 0.9149
• F1: 0.9146
• Precision: 0.9167
• Recall: 0.9149

Test Set

• Accuracy: 0.9241
• F1: 0.8316
• Precision: 0.8341
• Recall: 0.8291

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: 0.0002
• train_batch_size: 128
• eval_batch_size: 128
• seed: 42
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: cosine
• num_epochs: 5

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1	Preicision	Recall
0.5606	1.0	141	0.3018	0.9109	0.9107	0.9127	0.9109
0.2489	2.0	282	0.3185	0.9108	0.9089	0.9107	0.9108
0.1558	3.0	423	0.3378	0.9051	0.9028	0.9056	0.9051
0.0966	4.0	564	0.3472	0.9151	0.9149	0.9170	0.9151
0.0678	5.0	705	0.3693	0.9149	0.9146	0.9167	0.9149

Framework versions

• Transformers 4.52.3
• Pytorch 2.7.0+cu128
• Datasets 3.6.0
• Tokenizers 0.21.1