python - How to use tensorflow to build model with the parameters from a config file -
i have algorithm produce parameters feed tensorflow construct dl models. parameters : hidden_layer_neurons = [100, 200, 300, 400, 500].
because in tf, cannot use for-loop python, therefore, cannot use manner build multi-layer model reading elements in hidden_layer_neurons.
is there way solve question? in advance.
when constructing tf graph valid/possible allow kinds of python constructs loops. long can structure program in way modifying graph (using loops) not need happen during graph evaluation (e.g. run algorithm upfront , construct graph) there's no problem @ all.
if need iterate during graph evaluation (e.g. during session.run() ) you'd need flow control ops.
Comments
Post a Comment