---
license: apache-2.0
datasets:
- alibayram/turkish_mmlu
language:
- tr
base_model:
- google-t5/t5-small
---
# fine-tuned-t5-small-turkish-mmlu

<!-- Provide a quick summary of what the model is/does. -->

The fine-tuned [T5-Small](https://huggingface.co/google-t5/t5-small) model is a question-answering model trained on the [Turkish MMLU](https://huggingface.co/datasets/alibayram/turkish_mmlu) dataset, which consists of questions from various academic and professional exams in Turkey, including KPSS and TUS. The model takes a Turkish question as input and generates the correct answer. It is designed to perform well on Turkish-language question-answering tasks, leveraging the structure of the T5 architecture to handle text-to-text transformations.

### Training Data

<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->

@dataset{bayram_2024_13378019,
  author       = {Bayram, M. Ali},
  title        = {{Turkish MMLU: Yapay Zeka ve Akademik Uygulamalar 
                   İçin En Kapsamlı ve Özgün Türkçe Veri Seti}},
  month        = aug,
  year         = 2024,
  publisher    = {Zenodo},
  version      = {v1.2},
  doi          = {10.5281/zenodo.13378019},
  url          = {https://doi.org/10.5281/zenodo.13378019}
}


#### Training Hyperparameters

    learning_rate=5e-5
    per_device_train_batch_size=8
    per_device_eval_batch_size=8
    num_train_epochs=3
    weight_decay=0.01 


#### Training Results
 

![image/png](https://cdn-uploads.huggingface.co/production/uploads/669a700b990749decaab29af/xgl-5aCReHq8nA4RxgxhC.png)



#### Metrics

<!-- These are the evaluation metrics being used, ideally with a description of why. -->
Training loss was monitored to evaluate how well the model is learning and to avoid overfitting. In this case, after 3 epochs, the model achieved a training loss of 0.0749, reflecting its ability to generalize well to the given data.