What was the command/method used to make these quants?
#1
by Downtown-Case - opened
Just one, as an example.
I'm trying to quantize models myself in sd.cpp, but the filesize ends up different. To you specify certain tensors to be F16 or something?