Gibberish
#2
by
Niichanhaou
- opened
I get gibberish with this model. I tried both the base branch and TheBloke/EverythingLM-13B-16K-GPTQ:gptq-4bit-128g-actorder_True
I'm using ExLlama, any other llama-based models I run work totally fine. Only this one starts outputting things like
OOOOO​​\OO​OO​oOtt​OtOtOtO​ttt\​OOOtttOoot\oOOO​tOtOOtOtOtOtOtOtOtOtOtOtOtOtOtOtOtOtOtOtOtOtOtOtOtOtOtOtOtOtOtOtOtOtOtOtOtO
Any ideas?