All - need your help. I did upgrade of conda update -all. That messed up a lot of things.

My transfer learning classifier for inceptionv3 stopped worked. Previously, classifier was working really good - perfectly good results. Now after the upgrade which did a lot of libraries, it works but very poor recall and precision- as good as none.

Before, since early this year, it was perfect for months now. Except no upgrade. . I traced it to the issue that the imagenet may not be storing in the .keras\models. Not sure why it is not getting stored; where would I go to delete this and set this.

btw: this is windows 10; Anaconda setup.

use standard base model.

base_model = inception_v3.InceptionV3

using pre-existing standard weights

base_model = base_model(weights='imagenet', include_top=False)

No matter what i do it; i dont get this stored. Is it stored somewhere else and classifer when the model is compiled unable to find it?

Will be grateful, if there is may be a path i may have overlooked.