Was this trained on top of Llama 3.1 70B or Llama 3.3 70B?
#4
by ddh0 - opened
The blog post makes it seem like it was trained on top of L3.3, but the model card shows it as being tuned on top of L3.1:
It was trained on top of Llama 3.1 base model. (We used Llama 3.3 in the blog to compare against as a baseline.)
drishanarora changed discussion status to closed
that's pretty impressive