Quantisation
Quantisation reduces the numerical precision of a model's weights and calculations, for example from 16-bit to 4-bit numbers, so it uses less memory and runs faster.
What is Quantisation?
Quantisation lets large models run on laptops, phones and cheaper servers, usually with only a small loss in accuracy.
How it works
Values are mapped onto a smaller set of discrete levels, sometimes with calibration or retraining to preserve accuracy.
← All Artificial Intelligence & Superintelligence definitions · ← Back to the A–Z definitions