Tensorflow SavedModel - How to Iteratively Save

Question

I am using the new API SavedModelsince I found out that this is the “future” and should be preferred tf.train.Saver().

What I'm trying to accomplish is to save the model in each Nnumber of batches, and I want to save a maximum of 20 saved models. Obviously, I can control this on my own, but if I tf.train.Saver()can achieve this, I believe that a similar way can be done with SavedModel. I would also like to be able to explicitly save the model whenever I want, and that they all be called as model-001, model-002....

All this was possible with the help tf.train.Saver()- I could define max_to_keepand global_stepto save every era or so, and I could save when I want.

Is this possible using the API SavedModel? If so, how? An example will be appreciated.

It is reasonable to assume that it signature def mapwill be equal to all models.