sabanci-it-destek-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: 1.3585
• Accuracy: 0.8409
• F1: 0.8305

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: 3e-05
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 32
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 15
• mixed_precision_training: Native AMP
• label_smoothing_factor: 0.1

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1
No log	1.0	170	2.8936	0.4647	0.3646
No log	2.0	340	1.7200	0.7224	0.6580
2.6315	3.0	510	1.4805	0.7784	0.7387
2.6315	4.0	680	1.3844	0.7982	0.7736
2.6315	5.0	850	1.3450	0.8115	0.7967
1.1794	6.0	1020	1.3594	0.8078	0.7919
1.1794	7.0	1190	1.3484	0.8137	0.8032
1.1794	8.0	1360	1.3337	0.8373	0.8272
0.9009	9.0	1530	1.3531	0.8321	0.8234
0.9009	10.0	1700	1.3591	0.8402	0.8311
0.9009	11.0	1870	1.3585	0.8409	0.8305

Framework versions

• Transformers 4.57.3
• Pytorch 2.9.0+cu126
• Datasets 4.0.0
• Tokenizers 0.22.2