osjinc opened this issue 7 months ago · 0 comments
How about adding a GPT-3.5 Turbo Instruct model? With less censored responses?