Model soups failed until we fixed one line: our 26M grammar tagger from 11% to 1.9% false fixes

Our keyboard's grammar model is a 26.2M-parameter tagger we trained from scratch: one label per word...