Ignite: Avoid moving model to device in create_supervised*

[vfdev-5]

🚀 Feature

As discussed here and previous messages, according to pytorch docs on model and optimizer. User should not move model to another device after constructing an optimizer for the model:

• https://pytorch.org/docs/stable/optim.html#constructing-it
• https://github.com/pytorch/pytorch/issues/7844#issuecomment-503713840

We need to handle this in the code:

• remove the code that moves the model to device.
• update the docstring and add a warning note on changed behaviour, cite pytorch docs.
• update all basic examples : mnist etc who currently use unwanted behaviour.

cc @kai-tub

[kai-tub]

Hi, I would like to work on this and complete the work I've started. :)
When this issue is closed I will come back to my tiny_benchmark_wrapper.

[vfdev-5]

@kai-tub sounds good, thanks !

[kai-tub]

One quick question, what do you mean by:

... add a warning note on changed behavior

Is a normal Note: environment in the docstrings sufficient? Or do you mean warning from import warnings?

[vfdev-5]

Good question, I mean this

.. warning::

     Be careful

which is rendered as here: https://pytorch.org/ignite/metrics.html#ignite.metrics.EpochMetric