v1.0.4

Incorporate feedback from @neptunes5thmoon
Changelog:

• Fix bug in output size computation when crops were applied to the top level of the UNet
• Fix bug in residual layers - @neptunes5thmoon

v1.0.3

Incorporate feedback from @neptunes5thmoon

• warning is thrown if data being processed is smaller than the equivariant step size. Warning includes a basic description of possible side effects and a link to a relevant paper
• warning is thrown if ConvPass is initialized with Identity activation, explaining that this is used for test cases.
  • this may be changed in the future to allow normal weight initialization with the Identity transformation if that is something users want. We can always move the test time weight initialization into the test function
• NotImplementedError is thrown if the funkelab_api constructor is called with non default values for arguments we do not yet support.

new feature:

• residual layers on ConvPass

v1.0.2

add a LICENSE and py.typed files to be packaged with the library

v1.0.1

Fix variable name invariant_step to be equivariant_step

v1.0

Initial release