NMT 'test' script would benefit from an option to run the test on ClearML

[mmartin9684-sil]

Most of the NMT scripts that require a GPU have an option (--clearml-queue xxx) to send the job to a ClearML queue so that it can be executed on a shared GPU. The 'test' script does not support this command line option. This would be helpful to have since most models can't be run locally.

[mshannon-sil]

I see that the train.py script also does not include a --clearml-queue flag. The only part of the experiment pipeline that includes a --clearml-queue flag is translate.py, which was only added 7 months ago. @ddaspit would you like me to add --clearml-queue flags to both train.py and test.py, or should we direct users to specify --train and/or --test in experiment.py if they'd like to run on ClearML?

[ddaspit]

Does it work if you run the experiment script with just the --test flag?

[mmartin9684-sil]

Yes