Source of overhead in model.fit()

I have a code where I make repeated calls to model.fit()

(It's for implementing an A2C)

I've noticed that my first call to model.fit() has significant overhead, but not the calls afterwards.

So my question is what are the common causes of overhead for model.fit() and some potential fixes?

I'm using Tensorflow2.0 and python3.

https://stackoverflow.com/questions/66003063/source-of-overhead-in-model-fit February 02, 2021 at 10:59AM