Hello,
I was playing around with different models and eventually I ran into this issue.
It seems like every prompt starts at 6000 context tokens regardless of what is prompted. It’s not high priority for my current developments but I thought I’d report it here.
Great question! Llama is so much lower than the other models, it does break things. The trick is to go into the configure tab and reduce the token limits for input and output. You should be able to move them down below the limit and use the model! Sorry for the inconvenience.
