v0.5.0

Spectral normalization

v0.4.1

Major new feature: Jacobians of network outputs with respect to layer parameters.

v0.3.0

Main new features:

• Code generation
• Arbitrary loss gradients for OGD step
• File format changes (remove Input layer type)

v0.2.1

fix bug inside assert; check debug build on CI

v0.2.0

Adds gradient descent step on least-squares regression loss for one data point.

v0.1.0

beginning to stabilize

n/a

more work on PyPI

n/a

more work on getting uploads

n/a

another attempt to debug the CI publish-to-PyPI

v0.0.1 first release

testing release system and PyPI upload