v0.0.83

Fixed the pytorch-metric-learning version to 1.6.3.

v0.0.82

Features

domains can now be passed into CLossHook:

CLossHook(domains=["src", "target"])

v0.0.81

Features

• Added a new base validator SimpleLossValidator
• Added BNMValidator

v0.0.80

Features

• Added pretrained models for DomainNet126
• Added transforms.classification.get_timm_transform

Bug fixes

• Fixed bug where map_location wasn't being used in a useful way when downloading pretrained models.
• Fixed some formatting issues in the documentation.

v0.0.79

Features

• Added APValidator
• Added adapters.MultiLabelClassifier
• Added hooks.MultiLabelClassifierHook
• Added frameworks.ignite.IgniteMultiLabelClassification
• Added models.pretrained_scores

v0.0.78

Features

• Added VOCMultiLabel dataset
• Added Clipart1kMultiLabel dataset
• Added get_voc_multilabel dataset getter

Breaking changes

• Moved get_mnist_transform, get_resnet_transform, and GrayscaleToRGB to a new transforms module

v0.0.77

Features

Made DomainNet126 downloadable:

from pytorch_adapt.datasets import get_domainnet126
datasets = get_domainnet126(["clipart"], ["real"], folder=".", download=True)

v0.0.76

Code change

Simplified code in frameworks.ignite.checkpoint_utils by using the latest version of pytorch-ignite.

v0.0.75

Features

Removed memory-inefficient operation from MCCLoss

v0.0.74

Features

• Pass kwargs down from pretrained model getters to load_state_dict_from_url. For example, this allows map_location to be specified:

from pytorch_adapt.models import office31C

model = office31C(domain="dslr", pretrained=True, map_location=torch.device("cpu"))