crumpets.torch.utils module

class crumpets.torch.utils.Normalize(*args: Any, **kwargs: Any)

Bases: torch.nn.

forward(x)

class crumpets.torch.utils.Unpacker(*args: Any, **kwargs: Any)

Bases: torch.nn.

forward(sample, *_, **__)

crumpets.torch.utils.filter_state(own_state, state_dict)

crumpets.torch.utils.other_type(s)

crumpets.torch.utils.resume(path, model, optimizer)

Given parameters, extracts a training state, i.e. initializes a network and optimizer.

Parameters

• path – path to a pytorch snapshot (including model and optimizer states)

• model – a network architecture for that the extracted weights are applied to

• optimizer – an optimizer for which the extracted optimizer parameters are applied to

Returns

the loaded snapshot

crumpets.torch.utils.save(path, iteration, model, optimizer, **kwargs)

crumpets.torch.utils.try_dicts(k, *ds)

crumpets.torch.utils.try_types(k, *ds)