Glossary
What is GPTQ? Definition and meaning
Quantizes weights layer by layer to 4 bits or less, using approximate second-order information to limit error.
Definition
GPTQ: Quantizes weights layer by layer to 4 bits or less, using approximate second-order information to limit error.
In short: Post-training quantization using second-order information.
When it is used
Compressing language models for cheaper GPU inference.
Learn more
The full guide to GPTQ explains how it works, when to use it and how it compares with the alternatives, with examples. The glossary lists every term on the site from A to Z.
Previous term: GPT-6 · Next term: Gradient boosting