Pre-trained VAEGAN model uses older batch normalization implementation

pkmital / CADL

ARCHIVED: Contains historical course materials/Homework materials for the FREE MOOC course on "Creative Applications of Deep Learning w/ Tensorflow" #CADL

https://www.kadenze.com/courses/creative-applications-of-deep-learning-with-tensorflow/info

Apache License 2.0

1.48k stars 732 forks source link

Pre-trained VAEGAN model uses older batch normalization implementation #62

Open pkmital opened 7 years ago

pkmital commented 7 years ago

https://www.kadenze.com/forums/file-submission-generative-adversarial-networks-and-recurrent-neural-networks/threads/anyone-else-seen-this-issue-before

Should retrain a new model on the newer batch norm implementation in tf.contrib