Nx vs. Python performance for sentence-transformer encoding

Keep in mind you may not want to run eight instances. When I tried this, XLA took all cores and we could not push traffic enough to the serving. :smiley: