v2.8.1

Fixed some module import issues.

v2.8.0

Features

• Added the Datasets module for easy downloading of common datasets:
  • CUB200
  • Cars196
  • INaturalist 2018
  • Stanford Online Products

Thanks @ir2718!

v2.7.0

Features

• Added ThresholdConsistentMarginLoss.

Thank you @ir2718!

v2.6.0

Improvement + small breaking change to DistributedLossWrapper

• Changed the emb argument of DistributedLossWrapper.forward to embeddings to be consistent with the rest of the library.
• Added a warning and early-return when DistributedLossWrapper is being used in a non-distributed setting.
• Thank you @elisim!

v2.5.0

Improvements

• Allow scaling up the memory and batch size when using TripletMarginMiner
• Pull request: #689

Thanks @mkmenta !

v2.4.1

This is identical to v2.4.0, but includes the LICENSE file which was missing from v2.4.0.

v2.4.0

Features

• Added DynamicSoftMarginLoss. See PR #659. Thanks @domenicoMuscill0!
• Added RankedListLoss. See PR #659. Thanks @domenicoMuscill0!

Bug fixes

• Fixed issue where PNPLoss would return NaN when a batch sample had no corresponding positive. See PR #660. Thanks @Puzer and @interestingzhuo!

Tests

• Fixed the test for HistogramLoss to work with PyTorch 2.1. Thanks @GaetanLepage!

v2.3.0

Features

• Added HistogramLoss. See pull request #651. Thanks @domenicoMuscill0!

v2.2.0

Features

• Added ManifoldLoss. See pull request #635. Thanks @domenicoMuscill0!
• Added P2SGradLoss. See pull request #635. Thanks @domenicoMuscill0!
• Added the symmetric flag to SelfSupervisedLoss. If True, then the embeddings in both embeddings and ref_emb are used as anchors. If False, then only the embeddings in embeddings are used as anchors. The previous behavior was equivalent to symmetric=False. Now the default is symmetric=True, because this is usually what is done in self supervised papers (e.g. SimCLR).

v2.1.2

Bug Fixes

• Fixed bug where set_stats was not being called in TripletMarginMiner (#628)
• Made HierarchicalSampler extend torch.utils.data.Sampler instead of torch.utils.data.BatchSampler (#613)
• Made samplers documentation clearer (#615). Thanks @rheum !