turkish-hs-target-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: 0.2668
  • F1: 0.7805
  • Roc Auc: 0.8458
  • Accuracy: 0.7403

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: 48
  • 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 F1 Roc Auc Accuracy
0.4848 0.4348 100 0.4292 0.5819 0.7133 0.5283
0.3992 0.8696 200 0.3402 0.7118 0.7908 0.6311
0.3272 1.3043 300 0.3015 0.7529 0.8218 0.6857
0.3005 1.7391 400 0.2821 0.7684 0.8348 0.7124
0.2771 2.1739 500 0.2802 0.7749 0.8401 0.7251
0.2666 2.6087 600 0.2693 0.7746 0.8406 0.7263
0.2668 3.0435 700 0.2653 0.7766 0.8414 0.7365
0.2482 3.4783 800 0.2691 0.7715 0.8389 0.7213
0.2457 3.9130 900 0.2615 0.7767 0.8432 0.7359
0.2327 4.3478 1000 0.2616 0.7789 0.8443 0.7435
0.2299 4.7826 1100 0.2625 0.7768 0.8419 0.7302
0.2203 5.2174 1200 0.2612 0.7844 0.8480 0.7441
0.2164 5.6522 1300 0.2586 0.7859 0.8492 0.7505
0.2152 6.0870 1400 0.2649 0.7760 0.8422 0.7384
0.209 6.5217 1500 0.2635 0.7769 0.8424 0.7410
0.2085 6.9565 1600 0.2661 0.7819 0.8470 0.7371
0.2052 7.3913 1700 0.2651 0.7798 0.8451 0.7378
0.1983 7.8261 1800 0.2640 0.7860 0.8494 0.7467
0.1946 8.2609 1900 0.2666 0.7852 0.8489 0.7454
0.192 8.6957 2000 0.2667 0.7826 0.8474 0.7422
0.1965 9.1304 2100 0.2662 0.7804 0.8455 0.7416
0.1924 9.5652 2200 0.2668 0.7801 0.8458 0.7435
0.1854 10.0 2300 0.2668 0.7805 0.8458 0.7403

Framework versions

  • Transformers 4.49.0
  • Pytorch 2.5.1+cu124
  • Datasets 3.3.2
  • Tokenizers 0.21.0
Downloads last month
164
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-target-prediction

Finetuned
(20)
this model