Source code for super_gradients.common.data_types.enum.strict_load

from enum import Enum


[docs]class StrictLoad(Enum): """ Wrapper for adding more functionality to torch's strict_load parameter in load_state_dict(). Attributes: OFF - Native torch "strict_load = off" behaviour. See nn.Module.load_state_dict() documentation for more details. ON - Native torch "strict_load = on" behaviour. See nn.Module.load_state_dict() documentation for more details. NO_KEY_MATCHING - Allows the usage of SuperGradient's adapt_checkpoint function, which loads a checkpoint by matching each layer's shapes (and bypasses the strict matching of the names of each layer (ie: disregards the state_dict key matching)). """ OFF = False ON = True NO_KEY_MATCHING = "no_key_matching"