* Replaced TensorBoard logging with simple json file logging which removed the cumbersome TensorBoard and TensorFlow dependencies * Updated the Dockerfile for Python37 and PyTorch 1.4.0 * Updated the CI tests Py36 versions to Py37, also added a new CI test for PyTorch 1.4.0
1.1.1
------------------
* Fixed some versions in the requirements for TensorFlow and Pillow to avoid errors and segfaults * The module auto documentation has been updated for the new API changes
1.1.0
------------------
* A complete refactor of MemCNN with changes to the API * Factored out the code responsible for the memory savings in a separate InvertibleModuleWrapper and reimplemented it using hooks * The InvertibleModuleWrapper allows for arbitrary invertible functions now (not just the additive and affine couplings) * The AdditiveBlock and AffineBlock have been refactored to AdditiveCoupling and AffineCoupling * The ReveribleBlock is now deprecated * The documentation and examples have been updated for the new API changes
1.0.1
------------------
* Bug fixes related to SummaryIterator import in Tensorflow 2 (location of summary_iterator has changed in TensorFlow) * Bug fixes related to NSamplesRandomSampler nsamples attribute (would crash if no-gpu and numpy.int were given)
1.0.0
------------------
* Major release for completing the JOSS review: * Anaconda cloud and codacy code quality CI * Updated/improved documentation
0.3.5
------------------
* Added CI for anaconda cloud * Documented conda installation steps * Minor test release for testing CI build