3 ms·Because it’s fine-tuning an existing 7B-parameter language model, not training from scratch.by comex 4y agoBecause it’s fine-tuning an existing 7B-parameter language model, not training from scratch.