Back to Projects
// Project Case StudyApr 2025 – Jun 2025

GemmaTR

GemmaTR is my Turkish LLM fine-tuning project based on Google Gemma, Unsloth, and LoRA. I prepared a dataset of 400,000 Turkish Wikipedia entries and 50,000 law, education, and agriculture-focused QA pairs, trained four model variants on Google Colab, and published the work on Hugging Face.

Overview

GemmaTR is my Turkish LLM fine-tuning project based on Google Gemma, Unsloth, and LoRA. I prepared a dataset of 400,000 Turkish Wikipedia entries and 50,000 law, education, and agriculture-focused QA pairs, trained four model variants on Google Colab, and published the work on Hugging Face.

Problem

Turkish users have fewer open and specialized LLM resources than English users, especially for domain-focused question answering in areas like law, education, and agriculture.

Technical Approach

I prepared a large Turkish dataset, fine-tuned Google Gemma with Unsloth and LoRA, and iterated through four model variants during a 40-hour training process on Google Colab.

Result

I published the resulting model weights and training artifacts on Hugging Face, making the Turkish fine-tuning work reusable and inspectable by the community.

Related Writing

Breaking the Agglutinative Barrier: Custom NLP Pipelines for Turkish Summarization

Technologies

PythonGoogle GemmaUnslothLoRAHugging FaceTurkish NLPDataset PreparationGoogle Colab

External Links