3 ms·
Key question: is this an entirely new LLM trained from scratch, or is it fine-tuned on top of an existing foundation model like Llama 2 or Mistral? I think it'
by simonw 3y ago
Key question: is this an entirely new LLM trained from scratch, or is it fine-tuned on top of an existing foundation model like Llama 2 or Mistral?
I think it's a new foundation model, but the announcement doesn't make that clear to me.
- rockinghigh 3y agoThey are trained from scratch on potentially different datasets. Their architectures are very similar though.