Often it is far longer than writing it all yourself.

What I see as the advantage is taking doc comments or documentation and fine-tuning the engine on it.

Not embedding more words as numbers.

Not adding it to a prompt.

Literally fine tuning the model.

Does Alpaca or any LLaMA version let you do it? Cause Chat-GPT doesn’t.

And anyway I wouldn’t want to give our whole code over to some third company to re-use the way an artist’s work got jacked because it was publicly posted. I want to self-host the fine-tuned model.

THAT would be the killer feature, for businesses. The Web attracted businesses, not people. People moved over later.

Does Alpaca or any LLaMA version let you do it? Giving it (with its existing weights) a gigabyte of text files and letting it fine tune itself on that?