Resuming training is not reflected in Logger

[avik-pal]

When training is resumed using load_model, there is no change in the state of the Logger. So if we start to retrain at epoch x, the images and all step value will be stored starting again from 1.