1

I have few CNNs and an RNN built in TensorFlow, and would like to export them to HDF5 with minimal Keras implementation.

I don't mind adding a couple of lines of code to the end of my program which utilize Keras, but I cannot rebuild every one of my networks.

I can't seem to figure out how best to merge these, any insight?

SStensors
  • 71
  • 10
  • 2
    Welcome to SO. In order for others to be able to help you, please check 'How to create a Minimal, Complete, and Verifiable example' https://stackoverflow.com/help/mcve – desertnaut Jul 12 '17 at 17:34
  • The question is also why do you want to export them to HDF5? TensorFlow models can be exported to protobuf. You can even freeze those models to a single file. – Dat Tran Jul 12 '17 at 18:35
  • Well I want to export to HDF5 to have a universal file I can import back into Keras later and get weights, layers, etc without worrying about the initial model type – SStensors Jul 12 '17 at 19:09
  • As @desertnaut already mentioned, can you provide us with a Minimal, Complete and Verifiable example so that we can help you better. Thanks – Dat Tran Jul 13 '17 at 08:40
  • See this question: https://stackoverflow.com/questions/43402320/export-tensorflow-weights-to-hdf5-file-and-model-to-keras-model-json Right now the easiest way is to use Keras model – Maxim Sep 21 '17 at 17:34

0 Answers0