在R和python中运行Keras

Question

I am intrigued by the keras package, which allows running the well known deep learning library from within R. The benefits of running keras (and tensor flow and theano) without leaving R are obvious.

What I am unsure of is whether there is a possibility to share the intermediate result - ie a trained neural network - with colleagues on python workflow.

Answer 1

You could just save your model (architecture + weights + optimizer state) keras_save . And then your colleagues could load it from python . More information here .
Also, @Jindra Lacko noted, that keras_save is from kerasR - a different package than keras . Better solution - save_model_hdf5 function.