This CL automatically defuns the inner iteration of the eager Keras fit,...
This CL automatically defuns the inner iteration of the eager Keras fit, evaluate, and predict loops if a model supports defuning. A graph network is marked as supporting defuns if all of its internal layers also support being in a defun. Subclass models and custom layers must be manually marked as supporting defuns in the constructor, by setting `self._can_use_graph_functions=True` PiperOrigin-RevId: 219210710
Loading
Please sign in to comment