Why are the output texts all garbled?

Hardware & Software

CPU: 12700KF
GPU: 3090ti 24GB
OS: Win10 Pro 22H2
UI: text-generation-webui
CUDA: 12.1
PyTorch: Nightly
AutoGPTQ: v0.4.2+cu121

Model

Loader: AutoGPTQ(no_inject_fused_attention, disable_exllama)
Parameters Preset: LLaMA-Precise

Issue

Input:

Write a lambda function in Python that takes a list of strings as input and returns the element in the list whose first letter is 'J' or 'j'.

Output:

itionalziaition, import

importiveiveiveive own?itional Cob Uniti Censo_ daughiamsitionalisticingly

determin ##### <? Zür помеmundation millimeter Institutitionalitionalitionalitionalitional

itional Kontrolaifact for packageistiveoug importitionalive Censoitionalitionalitionalclipitionalitionalitionals

itionalvia listade

import ?rsitional importardiitionalitionalalitionalitionalitionalitionalitionalitional

ineitionalitionalсясяitional dependentitional {- circumst?itional import? {: importAAAA importdependent importces importSERTitional?

CHAPTER importive importitional

Anal <? <? SchiffdditionalConstraintsitionalitionalitionalwe import #!/itionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalivity <?itionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitional

Addition

I have also tested the same question to TheBloke/Phind-CodeLlama-34B-Python-v1-GPTQ, recieving the following output.

ziaitionalive importitional\istic divis variveitionalitionaliveiveitionaliveiveitionaliveiveitionaliveiveitionaliveiveitionaliveiveiveitionaliveiveitionaliveiveiveitionaliveiveiveitionaliveiveiveitionaliveiveiveiveitionaliveiveiveiveitionaliveiveiveitionaliveiveiveiveitionaliveiveiveiveitionaliveiveiveiveitionaliveiveiveitionaliveiveiveitionaliveiveitionaliveitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitionalitional

When it comes to TheBloke/CodeLlama-34B-Python-GPTQ, I got the following.

__ ______________________ =_____

Again, why the outputs are garbled? Any settings are wrong? I actually followed the instruction from your model card, and it seems that there are no specific settings.
By the way, the text-generation-webui is installed correctly, with official Llama-2-7b-chat-hf using fine.

Thank you so much!