yanekyuk/berturk-cased-keyword-discriminator

berturk-keyword-discriminator

This model is a fine-tuned version of dbmdz/bert-base-turkish-cased on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.4196
• Precision: 0.6729
• Recall: 0.6904
• Accuracy: 0.9163
• F1: 0.6815
• Ent/precision: 0.6776
• Ent/accuracy: 0.7365
• Ent/f1: 0.7058
• Con/precision: 0.6640
• Con/accuracy: 0.6151
• Con/f1: 0.6386

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: 16
• eval_batch_size: 16
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 8
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Precision	Recall	Accuracy	F1	Ent/precision	Ent/accuracy	Ent/f1	Con/precision	Con/accuracy	Con/f1
0.1899	1.0	1875	0.1927	0.6330	0.6682	0.9163	0.6502	0.6283	0.7398	0.6795	0.6438	0.5513	0.5940
0.137	2.0	3750	0.1988	0.6405	0.6959	0.9160	0.6671	0.6461	0.7475	0.6931	0.6297	0.6116	0.6205
0.101	3.0	5625	0.2375	0.6494	0.7188	0.9173	0.6824	0.6497	0.7743	0.7066	0.6488	0.6281	0.6383
0.0767	4.0	7500	0.2699	0.6533	0.7188	0.9154	0.6845	0.6575	0.7741	0.7111	0.6449	0.6285	0.6366
0.057	5.0	9375	0.3188	0.6696	0.6914	0.9163	0.6803	0.6790	0.7405	0.7084	0.6518	0.6112	0.6308
0.0423	6.0	11250	0.3646	0.6773	0.6846	0.9171	0.6809	0.6787	0.7388	0.7075	0.6746	0.5959	0.6328
0.0339	7.0	13125	0.4007	0.6711	0.6816	0.9151	0.6763	0.6782	0.7283	0.7023	0.6575	0.6055	0.6304
0.0282	8.0	15000	0.4196	0.6729	0.6904	0.9163	0.6815	0.6776	0.7365	0.7058	0.6640	0.6151	0.6386

Framework versions

• Transformers 4.19.2
• Pytorch 1.11.0+cu113
• Datasets 2.2.2
• Tokenizers 0.12.1