Live model logs

[nealmick]

Today the live server got an update for the training dashboard. When you click train model it will create a popup that shows the tensorflow logs from each epoch as the model is being trained live. Currently this works by requesting up to 10 lines per second, which i think can sometimes lag behind actual training so its possible to not see the end of training before redirect back to training page when finished. However there is also another button labeled logs, when clicked you should be able to bring back up all of the logs from training. This should make it much easier to track which model is which, and now can see how the model did for each epoch. Currently the default model has no logs, as it was trained before these updates, but it should also display a error message saying default model if the selected slot does not have a custom model.... or was trained before the update. Only new models trained after these updates will have logs saved. Once you reset the model slot the logs, along with the model weights and settings are all cleared and set back to default similarly to before. Currently these updates are only for the live server, I dont really plan on adding this to local installs, as you can just look at your terminal output and see the same thing.