turkish-hs-2class-prediction

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

  • Loss: 0.5990
  • Accuracy: 0.8748
  • Macro F1: 0.8696

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-06
  • train_batch_size: 16
  • eval_batch_size: 20
  • seed: 42
  • optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
  • lr_scheduler_type: linear
  • num_epochs: 10

Training results

Training Loss Epoch Step Validation Loss Accuracy Macro F1
0.585 0.1462 100 0.4524 0.7797 0.7752
0.4357 0.2924 200 0.3881 0.8099 0.8065
0.4058 0.4386 300 0.3497 0.8428 0.8337
0.3747 0.5848 400 0.3356 0.8583 0.8504
0.3571 0.7310 500 0.3340 0.8355 0.8318
0.3311 0.8772 600 0.3091 0.8665 0.8603
0.3065 1.0234 700 0.3085 0.8693 0.8614
0.3151 1.1696 800 0.2962 0.8684 0.8616
0.2795 1.3158 900 0.2885 0.8784 0.8722
0.263 1.4620 1000 0.3155 0.8647 0.8610
0.3116 1.6082 1100 0.3208 0.8629 0.8580
0.2744 1.7544 1200 0.3005 0.8803 0.8746
0.2868 1.9006 1300 0.3130 0.8675 0.8631
0.2454 2.0468 1400 0.3330 0.8611 0.8575
0.2262 2.1930 1500 0.3454 0.8629 0.8587
0.2483 2.3392 1600 0.3144 0.8757 0.8712
0.2451 2.4854 1700 0.3237 0.8647 0.8607
0.2384 2.6316 1800 0.3390 0.8638 0.8605
0.2347 2.7778 1900 0.3635 0.8611 0.8580
0.2683 2.9240 2000 0.3083 0.8748 0.8679
0.2047 3.0702 2100 0.3251 0.8748 0.8696
0.21 3.2164 2200 0.3381 0.8839 0.8784
0.1948 3.3626 2300 0.3383 0.8803 0.8754
0.1953 3.5088 2400 0.3495 0.8757 0.8707
0.1873 3.6550 2500 0.3539 0.8857 0.8787
0.201 3.8012 2600 0.3520 0.8784 0.8731
0.1935 3.9474 2700 0.3656 0.8629 0.8588
0.189 4.0936 2800 0.3486 0.8793 0.8737
0.1312 4.2398 2900 0.3845 0.8793 0.8737
0.1824 4.3860 3000 0.4035 0.8757 0.8696
0.1994 4.5322 3100 0.3820 0.8784 0.8737
0.1535 4.6784 3200 0.4042 0.8739 0.8683
0.1902 4.8246 3300 0.3990 0.8803 0.8730
0.1622 4.9708 3400 0.4224 0.8665 0.8619
0.1319 5.1170 3500 0.4311 0.8748 0.8694
0.1533 5.2632 3600 0.4505 0.8647 0.8609
0.1251 5.4094 3700 0.4523 0.8720 0.8670
0.1473 5.5556 3800 0.4535 0.8812 0.8762
0.1439 5.7018 3900 0.4566 0.8784 0.8727
0.1487 5.8480 4000 0.4472 0.8830 0.8772
0.1539 5.9942 4100 0.4414 0.8803 0.8743
0.1212 6.1404 4200 0.4778 0.8766 0.8720
0.1187 6.2865 4300 0.4734 0.8857 0.8806
0.1007 6.4327 4400 0.5087 0.8784 0.8731
0.1263 6.5789 4500 0.4983 0.8876 0.8826
0.1264 6.7251 4600 0.4999 0.8784 0.8737
0.1257 6.8713 4700 0.4943 0.8839 0.8776
0.121 7.0175 4800 0.5123 0.8775 0.8708
0.1012 7.1637 4900 0.5321 0.8775 0.8724
0.13 7.3099 5000 0.5426 0.8748 0.8703
0.1061 7.4561 5100 0.5380 0.8784 0.8732
0.1028 7.6023 5200 0.5481 0.8739 0.8685
0.1236 7.7485 5300 0.5456 0.8803 0.8740
0.0889 7.8947 5400 0.5653 0.8784 0.8722
0.0912 8.0409 5500 0.5781 0.8748 0.8699
0.1035 8.1871 5600 0.5711 0.8793 0.8736
0.109 8.3333 5700 0.5692 0.8793 0.8729
0.0996 8.4795 5800 0.5694 0.8793 0.8736
0.1158 8.6257 5900 0.5886 0.8720 0.8670
0.1008 8.7719 6000 0.5973 0.8702 0.8660
0.0859 8.9181 6100 0.5815 0.8803 0.8746
0.0927 9.0643 6200 0.5840 0.8766 0.8711
0.0918 9.2105 6300 0.5862 0.8775 0.8724
0.0661 9.3567 6400 0.5912 0.8784 0.8732
0.0884 9.5029 6500 0.5923 0.8784 0.8730
0.0845 9.6491 6600 0.6021 0.8748 0.8698
0.1071 9.7953 6700 0.6044 0.8748 0.8698
0.1016 9.9415 6800 0.5990 0.8748 0.8696

Framework versions

  • Transformers 4.49.0
  • Pytorch 2.5.1+cu124
  • Datasets 3.3.2
  • Tokenizers 0.21.0
Downloads last month
80
Safetensors
Model size
112M params
Tensor type
F32
·
Inference Providers NEW
This model is not currently available via any of the supported Inference Providers.

Model tree for HrantDinkFoundation/turkish-hs-2class-prediction

Finetuned
(20)
this model