Does not work. The model size is wrong too. 1.5B x 5.5BPW should be ~1.6GB
#1
by
imoc
- opened
So...?
Oh, I don't know
I used AutoQuants to quantize it, which himself uses turboderp/exllamav2 to quantize. So I don't know.
I will try to requantize it
How did you calculate this "1.6GB" ?