Possibility to give callback functions (for step and epoch finished) to method 'trainNetwork'?

[fjug]

This would make it simpler to 'observe' some properties during training, maybe even export infos required to use TensorBoard on models that are currently trained? (Not sure about the last one...)