single mention · inference & serving
Abs-max quantization
Quantization that uses absolute-maximum scaling and is reported sufficient with appropriate calibration.
The sources on this page name it without explaining how it works. The line above is the extraction's one-line gloss. See inference quantization for the methods filed beside it.
Also called Abs-max quantization with calibration, naive abs-max quantization.
Filed under inference & serving › inference quantization