Quantization Hyperparameters Optimization is an optional feature that enables the use of one of optimizers to automatically find the optimal hyperparameters for quantization. The definition of optimal depends on the optimizer and its configuration and may not necessarily be the best solution, for example, an optimizer may use only an approximate algorithm to search for optimal hyperparameters.
An optimizer uses hyperparameters metadata returned by quantization algorithms to create search space from which it chooses hyperparameters for algorithms. The configuration of hyperparameters metadata generation is done at algorithm level. For details, see description of a specific quantization algorithm.
Currently, only the following optimizers are available:
