Minisforum AI X1 Pro: Gerbil – Run Large Language Models Locally

This is a series of articles looking at the Minisforum AI X1 Pro running Linux.

Unlike many mini PCs, the AI X1 Pro has a very capable integrated GPU. I want to see what performance is like running machine learning software on the Radeon 890M.

In this article I’m looking at Gerbil, a desktop app designed to let you run Large Language Models locally with the minimum of fuss. The software is powered by KoboldCpp which itself is a highly modified fork of llama.cpp.

Installation

I evaluated Gerbil with the latest release of their AppImage downloaded from their GitHub repository. At the time of writing that’s Gerbil 1.4.3.

Make the file executable with the command:

$ chmod u+x Gerbil-*.AppImage

When trying to run the AppImage, you may get an error along the lines of

This may confuse newcomers as there isn’t actually a file /tmp/.mount_Gerbil5zECMd/chrome-sandbox.

My recommended way to resolve this type of issue is to create a file /etc/apparmor.d/gerbilappimage (e.g. sudo nano /etc/apparmor.d/gerbilappimage). with the following lines:

You’ll need to edit the profile line to where you’ve saved the AppImage.

Then issue the command:

$ sudo systemctl reload apparmor.service

Next page: Page 2 – Running Gerbil

Pages in this article:
Page 1 – Introduction
Page 2 – Running Gerbil

Complete list of articles in this series: