Pleas Stop Training Giant Language Models

i like the idea of small models for longer. Sounds like I could train some limited set on my own little mac and not have to pay openia any api fees!