Boris Power on X: "@minafahmi_ @OpenAI @goodside This happens with all the models in our API when there’s a tiny difference (&ampampampampampampampampampampampampampampampampampampampampamplt;1%) in probability between the two top tokens, due to non determinism. Once you get one different token then the completions might start to diverge more" / X - discu.eu

Linking pages

Non-determinism in GPT-4 is caused by Sparse MoE - 152334H https://152334h.github.io/blog/non-determinism-in-gpt-4/ 181 comments

Related searches:

Search whole site: site:twitter.com

Search title: Boris Power on X: "@minafahmi_ @OpenAI @goodside This happens with all the models in our API when there’s a tiny difference (&ampampampampampampampampampampampampampampampampampampampampamplt;1%) in probability between the two top tokens, due to non determinism. Once you get one different token then the completions might start to diverge more" / X

See how to search.

Submit link to: