cerebras.modelzoo.config_manager.config_classes.base.sparsity_config.SparsityAlgorithmConfig#

class cerebras.modelzoo.config_manager.config_classes.base.sparsity_config.SparsityAlgorithmConfig(**kwargs)[source]#
algorithm: Literal['static', 'gmp', 'set', 'rigl'] = 'static'#

Sparsity Algorithm to apply. [“static”, “gmp”, “set”, “rigl”]

params: dict = <object object>#

Parameters such as init_method, sparsity, param_name_patterns, schedule etc These vary based on the type of sparsity chosen. Please see sparsity documentaion for valid parameters for each type