News for hackers. Stuff that matters.

A comment on DeepSpeed Chat: Easy, fast and affordable RLHF training of ChatGPT-like models. Parent comment

Re:DeepSpeed Chat: Easy, fast and affordable RLHF training o... by UncleEntity on Wednesday April 12, 2023 @08:59PM (#35549437)
For the 1.3 billion parameter one they do mean train from scratch on your 2 hour “coffee break”.

Which totally begs the question, Microsoft AI researchers get two hour coffee breaks?

—edit—

The way they word it is confusing but, yeah, fine tune the model on your two hour coffee break.

[ Parent | on HN ]
Re:DeepSpeed Chat: Easy, fast and affordable RLHF training o... by MasterScrat on Wednesday April 12, 2023 @09:05PM (#35549480)
No, you can see the time breakdown in the table under the "coffee break" quote: it is the time for the 3-step RLHF process only. Training a 1.3B parameter model from scratch is still a very large undertaking.
[ Parent | on HN ]