8k on this model
#5
by
rombodawg
- opened
Is it possible to get 8k token context on the 65b guanaco model with the new methods?
In theory yes. There's no SuperHOT 65B LoRA yet so I'm not able to make a 65B 8K model yet. But new techniques are being worked on which should work with any model, even without new training/LoRA applied. So check back in a few days and it may well be possible
Sweet thanks for the reply