Tool loops (and literal sentence loops) #25
Description
Are you aware of this issue?
I've seen posts about this in other issues trackers, on reddits, etc. - it seems like a common problem.
I keep wondering if it's an issue with the inference software, or something related to training, e.g. (for tool loops) the model failing to recognize failure signals?
Though in the case of a single sentence repeating over and over, within the same turn, I'd expect this is more likely an issue with the inference engine running the model and failing to detect repetition?
(This is an incredible model for it's size and speed, btw! I hope it improves even more with future updates. 😊)