Closed SabinStargem closed 1 year ago
Who converted this model? n_gqa should be 8 for this model I think.
Venketh. They are probably still learning the ropes, I am guessing? Anyhow, I reported your solution on their model repository. With any luck, this would be fixed up in a jiffy.
https://huggingface.co/venketh/WizardLM-1.0-Uncensored-CodeLlama-34b-GGUF/tree/main
I am not entirely sure as 34B was only added upstream recently, and I have not tried running it myself. Lets see what they say first
I tried out The Bloke's version of the model, and it works fine. I am closing this issue, since it was a mishapen model and not a KoboldCPP problem.
As the title mentioned, I tried out WizardLM 34b, and it didn't boot up.