# LoRA + QLoRA: Parameter-Efficient Fine-Tuning Devrim — Hu 2021'den Dettmers 2023'e

> Source: https://sukruyusufkaya.com/learn/llm-muhendisligi/lora-qlora-parameter-efficient-finetuning
> Updated: 2026-08-22T08:40:33.090Z
> Category: LLM Mühendisliği
> Module: Modül 14: Fine-tuning — SFT, LoRA, QLoRA
**TLDR:** LoRA (Hu 2021): low-rank decomposition fine-tuning — base weights frozen, sadece küçük adapter eğit. %1 parameters, %95+ quality preservation. QLoRA (Dettmers 2023): 4-bit base + LoRA, 70B model'i consumer GPU'da fine-tune. NF4 quantization, paged optimizer. Türkçe pratik: $5K maliyetle production Türkçe Llama-3 70B.