Quick question about bagel
#1
by
algorithm
- opened
Hi jondurbin,
Thank you very much for these models and your amazing datasets.
Quick question, I noticed in your readme you wrote that tinyllama isn't really a useful base model.
So I was wondering if you've considered using phi-2 instead, as base model?
It's a surprisingly capable model.
No pressure of course, just a suggestion :)
Thanks!