v1.0.1 DeepSaki - Patch Release

What's Changed

• Update documentation by @sascha-kirch in #26
• Update logo font. add social card. added meta tags. by @sascha-kirch in #29
• Add dev status identifier by @sascha-kirch in #33
• Increase version to 1.0.1. Changed refferences for social cards. by @sascha-kirch in #34
• Update changelog. by @sascha-kirch in #37

Full Changelog: v1.0.0...v1.0.1

v1.0.0 DeepSaki - Major Release

Changes:

• Framework:

  • Upgrade from python 3.8 to python 3.10 since it supports a wide range of tensorflow version(2.6-2.14)
  • update of project structure to a modern SW project
  • automated CI/CD
  • docstrings and type annotations
  • initial automated documentation using mkdocs and mike supporting multiple version. Will be further improved in a future.
  • automated code coverage reports
  • automated publishing of packages on release
  • add changelog and contribution documentation
  • providing a dockerfile for the dev environment

• DeepSaki.initializers:

  • renamed DeepSaki.initializer -> DeepSaki.initializers
  • Refactored function make_initializer_complex() into a class called ComplexInitializer

• DeepSaki.layers:

  • renamed ResidualIdentityBlock -> ResidualBlock
  • New Layers:
    • LearnedPooling
    • FFT3D
    • iFFT3D

• DeepSaki.losses:

  • New abstract base class for image based losses.

• DeepSaki.optimizers:

  • renamed module from optimizer -> optimizers
  • new method switch_optimizer() to change the current optimizer of SwatsAdam or SwatsNadam optimizer.

• DeepSaki.augmentations:

  • renamed module from regularizations to augmentations since it only contained augmentations.

• DeepSaki.utils:

  • renamed DetectHw() -> detect_accelerator()
  • renamed EnableXlaAcceleration() -> enable_xla_acceleration()
  • renamed EnableMixedPrecision() -> enable_mixed_precision()

• New modules:

  • DeepSaki.types: provides type definitions and enums used in the package.

• Misc:

  • general code cleaning and refactoring.
  • changed print expressions to logging
  • using pyproject.toml for project config

• Outlook:

  • improved changelogs with issue tickets
  • new modules are planned, e.g. a diffusion module and a framework module featuring CycleGAN and DDPM.

DeepSaki v0.1.3

Fourier Convolution does now consider circular convolution

Full Changelog: v0.1.2...v0.1.3

DeepSaki v0.1.2

Release Note

• Added layers working with fourier transformations
• code cleaning

v0.1.0

Initial release v0.1.0