generating weird answers

[rpking1107]

first of all, thanks a lot for the amazing project. My GPU is 3060 12gb and cant run the 13b model, viand somehow oobabooga doesnt work on my CPU, Then i found this project, its so conveinent and easy to deploy!

But when i tried this on my computer i got really weird answers from models (also quite slow)
The model im using is this one: https://huggingface.co/anon8231489123/gpt4-x-alpaca-13b-native-4bit-128g/tree/main/gpt4-x-alpaca-13b-ggml-q4_1-from-gptq-4bit-128g. and here are the running time and response i got, these are clearly wrong, did i do anything wrong here? thanks!

[LostRuins]

Try formatting the prompt in the Alpaca format, or use the Instruct mode from the UI.