2 ms·Could be prompt based memory, or fine-tuning a small part of the big model.by visarga 4y agoCould be prompt based memory, or fine-tuning a small part of the big model.