Claude has a nice trick where you prefill the start of the assistant response, and it continues from there. Anyone know if Llama 3 can do the same thing?
Looks like the answer is yes :D x.com/charliebholtz/…
Looks like the answer is yes :D x.com/charliebholtz/…
@jeremyphoward Can’t you do this manually with any LLM? Are you asking about whether a particular inference codebase supports it?
@jeremyphoward @jeremyphoward That sounds like a cool feature Llama 3 can also pick up where you left off.
@jeremyphoward I spotted that ability for Reka too docs.reka.ai/guides/001-cha…
@jeremyphoward Same trick works great for GPT-4, GPT-3.5, too!
@jeremyphoward This works for me (just have to use the quite ugly prompt template): replicate.com/p/ba3xq5b3k9rg…
@jeremyphoward claude told me its unethical to impersonate an ai assistant in this way
@jeremyphoward They actually already do that in the official repo :) github.com/meta-llama/lla…