I think these guys could benefit from better prompts. When I use chat prompting… - system: sets the context for the bot - “You are…” - “Act like…” - “P

I think these guys could benefit from better prompts. When I use chat prompting… - system: sets the context for the bot - “You are…” - “Act like…” - “Pretend…” - assistant: specifies the task to be performed - “Classify…” - “Extract entities…” - “Translate…” - user: asks the question, adds examples These guys were asking questions and giving examples in the assistant prompt. I feel like that messed up the LLMs responses. I can’t wait to try Vicuña and MPT. I can’t quite figure out how to host them in Azure without an expensive VM. Maybe Azure DataBricks…?