Fine-tuning is the practice of adapting a pre-trained model to a particular task or domain using a relatively small, targeted dataset. Full fine-tuning of every weight is now mostly replaced by parameter-efficient approaches like LoRA and QLoRA, which deliver similar quality with a fraction of the GPU memory. Done well, fine-tuning bakes in domain-specific language, tone, format and task behaviour; done poorly, it can break the model's underlying abilities — the classic 'catastrophic forgetting' failure. In most production scenarios it is the heavier hammer reached for after Prompt Engineering and RAG have been tried.
MEVZU N°124ISTANBULYEAR I — VOL. III
Glossary · Intermediate · 2018
Fine-tuning
Adapting a pre-trained model to a specific task using smaller, targeted data.
- EN — English term
- Fine-tuning
- TR — Turkish term
- İnce Ayar (Fine-tuning)