Spectral Normalization for tf.keras. This repo contains minor adjustments to the Keras implementation SpectralNormalizationKeras (https://github.com/IShengFang/SpectralNormalizationKeras) to work with tensorflow.keras.
This is useful, because some TF functionalities (like efficient data loading) only work with tf.keras, and tf.keras and keras layers cannot be combined.
- All input_dim are casted to int (See this pull request: IShengFang/SpectralNormalizationKeras#4).
- Imports are changed from keras to tf.python.keras.
- Clone the SpectralNormalizationKeras repository (https://github.com/IShengFang/SpectralNormalizationKeras)
- Replace SpectralNormalizationKeras/SpectralNormalizationKeras.py with the corresponding file in this repository (SpectralNormalizationTensorflowKeras/SpectralNormalizationKeras.py)