CODEWITHSUNDEEP
variablessessiontensorflow
Lay González
Lay González

Reputation: 2985

In tensorflow, can I construct multiple times a graph and automatically reuse the same variables in all the graphs?

I might be wrong, but it’s my understanding that as long as I’m using the same session, it doesn’t matter how many times I build a graph, the variables will be the same, so if I train them for one graph, the next time within the same session I build the graph, the variables will already be trained. Am I wrong? Perhaps I need to use scopes? Or do I need to manually copy the variables between graphs?

Upvotes: 2

Views: 173

Answers (1)

benjaminplanche
benjaminplanche

Reputation: 15119

tf.make_template() may be what you are looking for. It wraps a provided set of tensor operations so that it does variable sharing.

Upvotes: 1

Related Questions