Added --gpu flag to run model on a gpu

[mutherr]

I noticed there wasn't a way to run the model on a gpu, so added a flag to set the device for the pipeline.

[rwetzold]

This is great! It will throw an error on my side though if run without the --gpu flag

UnboundLocalError: cannot access local variable 'device' where it is not associated with a value

[mutherr]

Resolved!

[rwetzold]

Works nicely! Now we just need to get this merged somehow

[mutherr]

It says something about workflows requiring approval. Maybe related to merging?

[mutherr]

Oh, no you might have to merge it, since you have write access.

[rwetzold]

I don't, we need @simonw

[ecentinela]

@simonw is this going to be merged?