v0.7.0

This release changes the order in which channels get loaded within the BENv2 Dataset and is now consistend with the order of S2 documentation.

v0.6.9

Improving messaging and implement messages in BENv2 (#117)

* Improving messaging and implement messages in BENv2

* adding optional timestamp to messages
moving BENv1 to messages

* remove old tst

* version update

v0.6.4

Updating references so that the documentation is build correctly (#110)

* Updating references so that the documentation is build correctly

adding drop path as its own parameter instead of re-using dropout

* increasing version number

v0.6.1

making ilmconfiguration a real dataclass (#106)

v0.6.0

This release adds dataloading in preparation for BENv2 and default metrics for multi-label classification

v0.5.1

This release relaxes the dependencies for torch. Every version of torch 2 can now be used except for versions 2.0.1 and 2.1.0, which had issues with their dependencies.

v0.5.0

This version unifies the interface of all data-loading-related classes. It introduces the throughput test data set to test how fast your model actually is without having to wait for data loading.

v0.4.10

There are fewer dependencies now in the base version of configilm.
The full version can still be installed with configilm[full] (which supports data modules etc.), and a smaller version can be installed with configilm[bigearthnet] (which adds support for lmdb reading of bigearthnet files).

By default, only the base version is installed which does not support all features in the .extra submodule.

REMARK: make sure to use one of the python versions 3.9 - 3.11 to be able to install the right torch version.

v0.4.9

This release adds fusion methods in general as well as their respective tests. No documentation of the methods is added yet.

v0.4.8

This release fixes a bug in the RSVQA LR and HR dataloader that try to load inactive answers from the json files