add "no logging" option

[CarloLucibello]

I may be wrong, but I see no way to entirely avoid logging during training, which sometimes may be convenient for quick exploratory experiments.

I suggest to have

trainer = Trainer(logger=None) 

construct a trainer that does no logging at all

[CarloLucibello]

wait, actually this avoids logging

trainer = Trainer(logger=False, checkpoint_callback=False)

If checkpoint_callback is not set, a checkpoints folder is created in the root directory.

Two suggestions:

• both arguments should accept the None value in addition (or as an alternative) to False. It seems more natural to me and the first thing I attempted.
• Setting logger=False should also trigger checkpoint_callback=False.

If there is no support for these suggestions the issue can ble closed.

[awaelchli]

Setting logger=False should also trigger checkpoint_callback=False.

Maybe not. The user may want to save checkpoints but no logging turned on. The logger is for experiment tracking (loss curves and other visualizations) and independent of checkpointing.