single mention · inference & serving
Post-training calibration sample count
An implementation choice specifying calibration on 1,024 samples, without a named quantization mechanism.
The sources on this page only mention it in passing. The line above is the extraction's one-line gloss. See inference quantization for the methods filed beside it.
Also called 1,024-sample post-training calibration, calibrate the model using 1,024 samples.
Filed under inference & serving › inference quantization