CoreConfig¶

Class to configure the optimization process of the model:

class model_compression_toolkit.core.CoreConfig(quantization_config=None, mixed_precision_config=None, bit_width_config=None, debug_config=None)¶

A class to hold the configurations classes of the MCT-core.

Parameters:

• quantization_config (QuantizationConfig) – Config for quantization.

• mixed_precision_config (MixedPrecisionQuantizationConfig) – Config for mixed precision quantization.

• None (If) –

• used. (a default MixedPrecisionQuantizationConfig is) –

• bit_width_config (BitWidthConfig) – Config for manual bit-width selection.

• debug_config (DebugConfig) – Config for debugging and editing the network quantization process.