I'm surprised this is not available, seems like a critical feature for production use in my world, hope you guys can deliver it soon. Maybe in a minor release. This would also be nice for local usage because even my music freezes when I'm running topic modeling:)
On my laptop I have 4 cores 8 threads, in production GL will either run on 24 threads or 56. Once we grow I was planning to give it 128 cores if I see the need - but when I figured out the topic models perfomance I don't think this will be nessesary.