def _datasets_loader(r: registry.Registry):
# TODO add original torchvision and wrapped datasets (with kv dataset[i])
from catalyst_gan import datasets as ext_datasets
r.add_from_module(ext_datasets)
try:
from torchvision import datasets
r.add_from_module(datasets, prefix=["tv.", "torchvision."])
# wrapped datasets
datasets_to_add = _get_module_classes(datasets)
prefixes = ["tv.kv.", "torchvision.keyvalue."]
for name, cls in datasets_to_add.items():
cls_w = functools.partial(_TorchvisionDatasetWrapper,
base_dataset=cls)
r.add(**{f"{p}{name}": cls_w for p in prefixes})
except ImportError:
pass
def _models_loader(r: registry.Registry):
from catalyst_gan import models
r.add_from_module(models)
try:
from torch_mimicry.nets import sngan, ssgan, infomax_gan
r.add_from_module(sngan, prefix=['tm.', 'tm.sngan.'])
r.add_from_module(ssgan, prefix=['tm.', 'tm.ssgan.'])
r.add_from_module(infomax_gan, prefix=['tm.', 'tm.infomax_gan.'])
except ImportError:
pass # TODO: warning?
def _modules_loader(r: registry.Registry):
from catalyst_gan import nn
r.add_from_module(nn)
def _metrics_loader(r: registry.Registry):
from catalyst_gan import metrics
r.add_from_module(metrics)
def _batch_transforms_loader(r: registry.Registry):
from catalyst_gan import batch_transforms
r.add_from_module(batch_transforms)
def _callbacks_loader(r: registry.Registry):
from catalyst_gan import callbacks
r.add_from_module(callbacks)
def _criterions_loader(r: registry.Registry):
from catalyst_gan.nn import criterion
r.add_from_module(criterion)