This wraps up https://github.com/microsoft/IRNet for making predictions. Don't judge IRNet based on this, I haven't got a BERT-based model for it, haven't even set up a way to test multi-table queries, and have broken things in various ways. Watch this space though.

Steps:

  # Make sure sqlova container isn't running since this will use same port.
  # Make sure docker has access to >= 5GB of memory, and that you're not
  # short on disk space.

  make && make run

  curl -F "[email protected]" -F "q=how long is throgs neck" localhost:5050

  # You'll have to be very patient!  It takes a long time to load the model.
  # The first time you run this it will be especially long, since various
  # downloads and unpackings need to happen.  It wasn't practical to stick
  # all that in the docker container, at least not just yet.