ObservationType

class torch.ao.quantization.backend_config.ObservationType(value)

An enum that represents different ways of how an operator/operator pattern should be observed

OUTPUT_SHARE_OBSERVER_WITH_INPUT = 1

this means the output will use the same observer instance as input, based on qconfig.activation example: torch.cat, maxpool

OUTPUT_USE_DIFFERENT_OBSERVER_AS_INPUT = 0

this means input and output are observed with different observers, based on qconfig.activation example: conv, linear, softmax