Closed EndingCredits closed 9 months ago
Okay, I clearly didn't look hard enough, turns out I just neede to pass "skip_special_tokens": False,
into the API query.
I will leave this here as testament to my folly (and if anyone else is having similar issues). Feel free to mark as resolved.
Thank you for the solution! Closing now
Apologies for bringing up what is more an Ooba issue, but I am getting what seems to be incorrectly formatted outputs while querying the model through the text-generation-webui api with text completion.
Input prompt:
Returned Completion:
I did wonder if it might be the quants, but I've tried it with this model https://huggingface.co/LoneStriker/functionary-small-v2.2-6.0bpw-h6-exl2 as well as this one https://huggingface.co/LoneStriker/functionary-medium-v2.2-3.5bpw-h6-exl2 as well as the 4bpw gguf from https://huggingface.co/meetkai/functionary-small-v2.2-GGUF.
Any idea what could be causing this issue - it would be great to get this working with the text-generation-webui as it's just nice t have everything in one place rather than having to rely on multiple different backends.