16k context variant possible?
#2
by
InvictusCreations
- opened
First off: crazy good model, can't decide if this or Stheno is my favorite.
However it creates the longest responses of any llama3 finetune I've tried so far. (between 600 and 1000 tokens.) So the maximum default context is full within < 8 responses.
Any plans for extending the context to 12k or 16k?
My plan for the next version is Uncensored, 12k or 16k, to begin after the Uncensored version is completed. Due to my hardware limitations, I will do my best.
aifeifei798
changed discussion status to
closed