Trainable Gaussian Pool and np/tf flexible Padding

What’s Changed

• Gb/pad bug (#54) @grantbuster
• Gb/trainable pool (#53) @grantbuster
• adding rank attr to gaussian noise layer to enable input shape determ… (#52) @bnb32

New custom layers for sup3rcc precip

What’s Changed

• Gb/silu (#51) @grantbuster

Implemented gaussian average pooling layer

What’s Changed

• Gb/fix gpool (#50) @grantbuster
• Gb/gauss pool (#49) @grantbuster

Deprecating support for scikit-learn version < 1.2

OneHotEncoder sparse parameter renamed to sparse_output to conform with scikit-learn version >= 1.2

New custom layers

What’s Changed

• added custom functional layer for basic tensorflow functional layers … (#46) @grantbuster
• Bnb/fno (#45) @bnb32

Sup3r layer names and python versions

What’s Changed

• updated python requirement >= 3.8 (#44) @grantbuster
• missed super().init() call in concat/adder. Setting ._name instea… (#43) @bnb32
• input_feature names for concat/adder layers (#42) @bnb32

backwards compatibility for models trained with previous phygnn versions

What’s Changed

• some edits for backwards compatibility with older pandas and phygnn v… (#41) @bnb32

New depth to time feature

What’s Changed

• added depth to time spatiotemporal expansion option with tests (#40) @grantbuster

Fixed random seed function for early TF 2 versions

What’s Changed

• fixed random seed setting to be compatible with all tf 2 versions (#39) @grantbuster

New sup3r res topo input layers

What’s Changed

• Gb/wind topo (#38) @grantbuster