nomic-ai / gpt4all

gpt4all: run open-source LLMs anywhere

Home Page:https://gpt4all.io

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Meaningless strings for slightly bigger prompts when running on GPU

rds0001 opened this issue · comments

Bug Report

I am running GPT4All on a Windows Laptop with an NVIDIA GeForce GTX 1070 graphic card.
It runs quite fast and correct for small prompts like "Write a story". For bigger prompts, however that consist of several sentences but are still within the allowed context window, I get only output like "�y▅qlnkhwmgvdut▅ai|{xpo][cW`▅s\X_bjz{ZUeR{|{▅{y^QYOuwvTx|" etc. I do not have this error when running on CPU.

Steps to Reproduce

  1. Start GPT4All.
  2. Enter a prompt that consists of a small article summary (4-5 sentences).

Expected Behavior

It should provide the summary as output.

Your Environment

  • GPT4All version: Newest version as of 04/29/2024
  • Operating System: Windows 10
  • Graphic Card: NVIDIA GeForce GTX 1070
  • Chat model used (if applicable): mistral-7b-instruct-v0.1.Q4_0.gguf