3 ms·
Yeah, sorry for not being precise. We used the technique to fine tune around 40 GPT 1.5B models, including the chess one. It was very apparent that the techniq
by sillysaurusx 4y ago
Yeah, sorry for not being precise. We used the technique to fine tune around 40 GPT 1.5B models, including the chess one.
It was very apparent that the technique was working well. The kiss curve suddenly started dropping dramatically the first day we got it working.