README.md

metadata

library_name: transformers
license: mit
base_model: dbmdz/bert-base-turkish-cased
tags:
  - generated_from_trainer
metrics:
  - accuracy
  - f1
  - precision
  - recall
model-index:
  - name: turkish-sentiment3
    results: []
datasets:
  - winvoker/turkish-sentiment-analysis-dataset

turkish-sentiment3

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: 0.0876
• Accuracy: 0.9688
• F1: 0.9450
• Precision: 0.9525
• Recall: 0.9381
• Positive F1: 0.9713
• Positive Precision: 0.9650
• Positive Recall: 0.9776
• Neutral F1: 0.9981
• Neutral Precision: 0.9977
• Neutral Recall: 0.9985
• Negative F1: 0.8655
• Negative Precision: 0.8947
• Negative Recall: 0.8382

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	Precision	Recall	Positive F1	Positive Precision	Positive Recall	Neutral F1	Neutral Precision	Neutral Recall	Negative F1	Negative Precision	Negative Recall
0.2262	0.0290	200	0.1398	0.9485	0.9152	0.9084	0.9227	0.9525	0.9626	0.9426	0.9929	0.9886	0.9972	0.8003	0.7739	0.8285
0.1265	0.0581	400	0.1125	0.9590	0.9277	0.9369	0.9195	0.9626	0.9545	0.9709	0.9963	0.9961	0.9965	0.8241	0.8601	0.7910
0.1213	0.0871	600	0.1270	0.9545	0.9118	0.9573	0.8834	0.9593	0.9307	0.9896	0.9961	0.9953	0.9970	0.7801	0.9458	0.6637
0.0995	0.1162	800	0.1171	0.9606	0.9339	0.9284	0.9399	0.9636	0.9690	0.9582	0.9967	0.9979	0.9956	0.8415	0.8184	0.8660
0.108	0.1452	1000	0.0944	0.9667	0.9413	0.9513	0.9325	0.9697	0.9628	0.9768	0.9964	0.9935	0.9993	0.8578	0.8976	0.8214
0.0956	0.1743	1200	0.0944	0.9655	0.9418	0.9369	0.9470	0.9679	0.9738	0.9622	0.9976	0.9961	0.9991	0.8598	0.8408	0.8796
0.0971	0.2033	1400	0.0874	0.9688	0.9449	0.9533	0.9372	0.9713	0.9643	0.9784	0.9981	0.9978	0.9984	0.8652	0.8977	0.8349
0.0887	0.2324	1600	0.0876	0.9688	0.9450	0.9525	0.9381	0.9713	0.9650	0.9776	0.9981	0.9977	0.9985	0.8655	0.8947	0.8382

Framework versions

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