nntoolbox.utils.transfer module

Some utility functions for transfer learning

nntoolbox.utils.transfer.cut_model(model: torch.nn.modules.container.Sequential, sep: Callable[[...], torch.nn.modules.module.Module] = <class 'torch.nn.modules.pooling.AdaptiveAvgPool2d'>) → Tuple[torch.nn.modules.container.Sequential, torch.nn.modules.container.Sequential]

Cut a non-sequential model at the first instance of layer type

Parameters

• model –

• sep –

Returns

nntoolbox.utils.transfer.cut_sequential_model(model: torch.nn.modules.container.Sequential, sep: Callable[[...], torch.nn.modules.module.Module] = <class 'torch.nn.modules.pooling.AdaptiveAvgPool2d'>) → Tuple[torch.nn.modules.container.Sequential, torch.nn.modules.container.Sequential]

Cut a sequential model at the first instance of layer type

Parameters

• model –

• sep –

Returns