Add support for running NMT training stage on Hangfire
Currently, ClearML is required to train a NMT engine. If we add a Hangfire build job that executes the training stage of the NMT build, then an entire NMT build can be executed locally for development and testing purposes.
Would it spin up a docker image containing machine.py? Probably it would be better to have one master docker image that has both .net and python packages - and all repos sitting next to each other on the hard drive. Then, instead of just mounting the main hard drive, all the repos could be mounted (mount '..') and we could debug any of the code. This would be helpful to have an easy way to debug all aspects of the code.
I wasn't planning on spinning up a docker container. My plan is to use Python.NET to call Machine.py.
It would be nice to have a development docker container image in Serval that can hold all 3 repos and truly build it. If it needs to be a separate issue, let it be so.