I use llama for anything that falls under an NDA clause, e.g. data by which the client could be identified. Like asking it analyze existing code/algorithms, to document functions or parse data is done only locally.
if the local model got stuck I try to abstract the problem and present gpt with the general case (e.g. given 2 columns A,B, find... )
if the local model got stuck I try to abstract the problem and present gpt with the general case (e.g. given 2 columns A,B, find... )