# Quantization Matematiği: Symmetric/Asymmetric, Per-Tensor/Per-Channel/Per-Group, QAT vs PTQ

> Source: https://sukruyusufkaya.com/learn/fine-tuning-cookbook/ftc-quantization-mathematics-fundamentals
> Updated: 2026-08-21T22:56:09.437Z
> Category: Fine-Tuning Cookbook (Model-by-Model)
> Module: Part X — Quantization Engineering
**TLDR:** Quantization'ın matematiksel temeli: floating-point → integer mapping formülü, symmetric vs asymmetric quantization, per-tensor vs per-channel vs per-group granularity, QAT (Quantization-Aware Training) vs PTQ (Post-Training Quantization), bit-width seçimi. RTX 4090'da Llama 8B'nin 32 layer'ında her tensor'ün quantization karakteristiği.