turkish-sentiment

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: 0.0873
• Accuracy: 0.9685
• F1 Macro: 0.9442
• F1 Weighted: 0.9680
• Precision Macro: 0.9525
• Recall Macro: 0.9367

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: 64
• eval_batch_size: 128
• 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: cosine
• lr_scheduler_warmup_steps: 400
• training_steps: 1600

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1 Macro	F1 Weighted	Precision Macro	Recall Macro
0.2356	0.0290	200	0.1424	0.9503	0.9151	0.9497	0.9205	0.9102
0.1228	0.0581	400	0.1106	0.9580	0.9249	0.9570	0.9389	0.9130
0.12	0.0871	600	0.1209	0.9589	0.9233	0.9570	0.9529	0.9020
0.0965	0.1162	800	0.1099	0.9628	0.9344	0.9622	0.9464	0.9240
0.1107	0.1452	1000	0.0900	0.9676	0.9427	0.9671	0.9519	0.9345
0.0957	0.1743	1200	0.0925	0.9662	0.9426	0.9664	0.9385	0.9470
0.0978	0.2033	1400	0.0866	0.9686	0.9442	0.9681	0.9537	0.9358
0.0911	0.2324	1600	0.0873	0.9685	0.9442	0.9680	0.9525	0.9367

Framework versions

• Transformers 4.52.4
• Pytorch 2.6.0+cu124
• Datasets 3.6.0
• Tokenizers 0.21.2