General discussion and feedback.
#1
by
Lewdiculous
- opened
Heya! Is the recommended max --contextsize 8192 or do you believe it would have understanding of slightly higher contexts like 12288?
PS: Can I request the addition of a mascot waifu image for the model card? It is tradition :'3
Hey! Guess 8192k, this model gets very confused beyond that, unfortunately... About the card, that would perfect! I'll try to make some in the future for every model as well! =) Also, thanks for the Imatrix model!
@Endevor – This model is horny and I absolutely love that side :D
Will be looking forward for whatever else you're experimenting with. I tend to make the quants for myself anyways so might as well share too :')