3 ms·
It is as good as GPT-3 at most sizes. Instruct layer needs to be put on top in order for it to compete with GPT 3.5(which powers ChatGPT). It can be done with c
by terafo 4y ago
It is as good as GPT-3 at most sizes. Instruct layer needs to be put on top in order for it to compete with GPT 3.5(which powers ChatGPT). It can be done with comparatively little amount of compute(couple hundred bucks worth of compute for small models, I'd assume low thousands for 65B).
- stavros 4y agoThat's surprising to read, given that ChatGPT (at least the first version) was much worse than text-davinci-003 at following instructions. The new version seems to be much better, though.
- weird-eye-issue 4y agoNo, it wasn't