`neural_compressor.common.base_config`

Module Contents

`ConfigRegistry`
`BaseConfig`	The base config for all algorithm configs.
`ComposableConfig`	The base config for all algorithm configs.

`register_config`(framework_name, algo_name[, priority])	Register config decorator.
`get_all_config_set_from_config_registry`(...)
`register_supported_configs_for_fwk`(fwk_name)	Register supported configs for specific framework.

options

neural_compressor.common.base_config.register_config(framework_name: str, algo_name: str, priority: float | int = 0)[source]

Register config decorator.

The register the configuration classes for different algorithms within specific frameworks.

Usage example:: @register_config(framework_name=FRAMEWORK_NAME, algo_name=ExampleAlgorithm, priority=100) class ExampleAlgorithmConfig:

# Configuration details for the ExampleAlgorithm

Parameters:

framework_name – the framework name.
algo_name – the algorithm name.
priority – the priority of the configuration. A larger number indicates a higher priority, which will be tried first at the auto-tune stage. Defaults to 0.

class neural_compressor.common.base_config.BaseConfig(white_list: List[neural_compressor.common.utils.OP_NAME_OR_MODULE_TYPE] | None = DEFAULT_WHITE_LIST)[source]: The base config for all algorithm configs.

class neural_compressor.common.base_config.ComposableConfig(configs: List[BaseConfig])[source]: The base config for all algorithm configs.

neural_compressor.common.base_config.register_supported_configs_for_fwk(fwk_name: str)[source]

Register supported configs for specific framework.