Skip to content

Glossary: Transformers Basics

Quantization

Storing a model's weights in fewer bits, such as 4 or 8, to cut memory use.

Where the book teaches it

Related terms

FLOPs (floating-point operations)