▶ Quantization Labs

Quantization Labs

FP16/INT8/INT4, SmoothQuant, GGUF format, perplexity curves.

5Interactive labs
100%Single-file HTML
Interactive labs

All 5 labs in this category

Advertisement
LAB · 01

Quantization Calibration

See how calibration data choice affects quantized weight ranges.

Open lab
LAB · 02

GGUF File Format

Inspect a GGUF model file structure.

Open lab
LAB · 03

Perplexity vs Quantization Level

Curve of model quality (perplexity) as bit width drops.

Open lab
LAB · 04

FP16 vs INT8 vs INT4

Same weight tensor; see precision/range for each format.

Open lab
LAB · 05

SmoothQuant — Migrating Outliers

See how SmoothQuant moves difficulty from activations to weights.

Open lab