Add gradient norm to train output

[twuebi]

Printing the gradient norm alongside the training output can be helpful when tuning learning rates and for finding the gradient clipping threshold.

[danieldk]

The output is already a bit wide. I wonder if we really need the progress bar. It looks nice, but it takes up a lot of space and does not provide more information than just a percentage.

With the progress bar removed, we could add some more data points.

[twuebi]

I'm leaning towards already covered through Tensorboard. If we want to add it to the console: do we need a process bar or would a simple XY/YZ or XY% suffice?

[danieldk]

At least Tensorboard makes it possible to monitor a lot more things without things getting too crowded. In the terminal output I am typically only interested in the train/validation accuracy and sometimes loss. There already is a lot of visual information that make these kinda hard to spot. So now that you mention Tensorboard, maybe that is the best solution.

[twuebi]

That's already there for all models but the RNN.