does this output text?
#1
by
rombodawg
- opened
i tried this with the 6.7b models and it only repeats tokens, how was your luck?
This model isn't intended to be directly used - it's just 8 DeekSeep Coders in an MoE. It freaks out with any temperature over like 0.2.
So then what is the intended purpose of the model?
I used this as a training base to create Code-Conjurer-8x1.3B which is a finetuned version of this model. :)
I'm also not super happy with how that one turned out but it functions
Ok thanks for the information. It helps me out
rombodawg
changed discussion status to
closed