Settings?
What exactly settings to use for this model? I did everything you specified in the description and here too. That's terrible. It's repetitive and behaves strangely to the max.
I think I tested with 0.9 temp and 0.1 min_p. Realistically any system prompt should work, the one in that preset should work fine. Make sure you use the ChatML context and instruct template.
Those instructions should be updated, sorry :p
Repetitive behavior. Whatever settings I set, this model tries to repeat itself over and over, almost unusable for me. To be clear, I am using a gguf Q4_k_s. Temp 0.9, min_p 0.1, other neutral, no dry, ChatML cantext and instruct. Blank sys prompt.
Blank sys prompt.
I would highly not recommend using a blank system prompt with a model like this, it's almost entirely trained for system prompt-aligned roleplaying.
Also, the current Q4 quant is static, and likely isn't of the highest possible quality. I'm working on imatrixed ones right now ^-^ but it may be a while still