Simon Gardling
Simon Gardling
Why?
I can also reproduce this error. But I only get this error on models that would require more ram than my gpu can provide. LLaMA-7B and LLaMA-13B work, but LLaMA-30B...
doesn't even work
You're using windows.
I'm able to reproduce the performance improvements, impressive work!
I'm ok, thanks.
I pulled this branch and figured out how to build it (I dont know anything about android development). Great work everyone. I will keep silently lurking.
> Okay, I'd appreciate a short round of testing, ideally not just the Quick Connect stuff. Either way, I'll probably merge this into `redesign` some time tomorrow... Is there something...
I'll compile it right now and test then!