updates

move search parameters to settings / add genetic algorithm settings

stopping condition on training or validation dataset

add multidevice test, fix imports

add multidevice test, fix imports

fix layer scaling bug

add class weight param to training configuration

add class weight param for training

workaround for BatchNormalization uninitialized variables

fix imports

pypi issue

version++

fix refactoring

setup tf session before loading model

Merge pull request #16 from guybedo/feature/multi_device_model