4B model is the same size as 1.7B model
#2
by
Flexan - opened
They don't have the same hash, but they have the exact same file size and parameter count.
Apologies -- we've been messing a lot with the models lately, deprecating older ones, and so a couple buggy releases have occurred. The newer model(s) (both param variants) will now produce reasoning and risks xml structure. We've trained with a larger corpus and this CoT approach provides more precision and recall. Still iterating!
Flexan changed discussion status to
closed