Detailed Guide to Quantisation Methods for LLMs

Author(s): Parth Chokhra

Originally published on Towards AI.

A Visual Step-by-Step Guide to Popular Quantisation Techniques

Quantisation is the process of reducing the precision of numbers used in a model; for example, storing weights in 8-bit integers instead of 16- or 32-bit floats. This makes models smaller, faster, and more efficient to run, often with only a small drop in accuracy. For large language models, quantisation is especially important because of their size and hardware demands. To understand this, look at the example below.

The article explains quantisation techniques used in machine learning, emphasizing the significance of Quantisation Aware Training (QAT) and Post-Training Quantisation (PTQ). It discusses various methods for quantising large language models, offering insights into how these approaches can reduce model size and improve efficiency, along with a detailed breakdown of different quantisation strategies and their applications, as well as the file formats used for storage and deployment of quantised models.

Read the full blog for free on Medium.

Join thousands of data leaders on the AI newsletter. Join over 80,000 subscribers and keep up to date with the latest developments in AI. From research to projects and ideas. If you are building an AI startup, an AI-related product, or a service, we invite you to consider becoming a sponsor.

Published via Towards AI