News for hackers. Stuff that matters.
Re:Meta Llama 3 by cjbprime on Thursday April 18, 2024 @10:59PM (#40083061)
Oh cool! But at the cost of twice the VRAM and only having 1/8th of the context, I suppose?
Re:Meta Llama 3 by modeless on Saturday April 20, 2024 @02:49AM (#40095385)
Llama 3 70B takes half the VRAM as Mixtral 8x22B. But it does need almost twice the FLOPS/bandwidth. Yes, Llama's context is smaller although that should be fixable in the near future. Another thing is that Llama is English-focused while Mixtral is more multilingual.