HN
Hacker News
Top
New
Best
Ask
Show
Jobs
Comment by bestcoder69 | Hacker News Reader
Parent
Full thread
bestcoder69
·
What instruction tuned LLM is better?
View on HN
yunyu
·
FLAN-UL2
MacsHeadroom
·
Not per bit or per watt. LLaMA-30B is 16GB and draws 40 watts from a 4090 GPU.
yunyu
·
LLaMa isn't instruction tuned
Reply on news.ycombinator.com