Hardware components | ||||||
 |
| × | 1 | |||
 |
| × | 1 | |||
Software apps and online services | ||||||
 |
| |||||
| ||||||
| ||||||
In this project we install the ollama chat application and download the Deepseek open reasoning model.
Jetson NanoA souped-up Raspberry Pi computer. I purchased this Jetson Nano to take Nvidia on-line training: Getting Started with Artificial Intelligence on Jetson Nano. My engineering society accepted this course as 8 hours of professional training.
Set up the hardware according to Nvidia's Get Started with Jetson Nano Developer Kit. My board has a yellow jumper on the J48 power header and runs from a laptop charger.
The wireless, fan and camera are options added to the basic SBC single board computer. Ordinary chargers are limited to the slower, 2 core 5Watt mode.
Download the zipped SD Card Image from the Nvidia Developer's site. Image 4.5.1 from 2021 is still the most recent.
You can use a larger drive. Use Balena Etcher to burn the image to the card. Insert the SD card into the flash reader on your PC or use an adaptor.
Click Flash from file and navigate to the zipped download file. Etcher extracts sd-blob-b01.img from the compressed file automatically.
Select the target drive. Important! We are about to erase a storage drive. Be careful not to select a drive on your PC.
This is important. You can damage your PC computer if you are not careful. Check the image and storage device you are about to Flash!
This takes a while. Twenty minutes for my choice of 32GB SD card. Click skip verification to save a few minutes. You may now eject the card and insert it into the Jetson Nano compute module.
Insert SD CardLocate the SD slot on the carrier board, insert the card, tweezers may help. The metal pins on the SD face the board. There is a clicky spring.
Like the Raspberry Pi this is the drive for your operating system and your /home folder files.
First BootLike many Linux distros you will be asked to have username and passwords ready. This first boot decompresses the image and we end up with a normal desktop. You should explore the default applications.
Open a terminal and type df for free disk space. We still have over half our SD card free for files and applications.
Connect to the network and do sudo apt update. Full upgrade is recommended but not necessary. Progress may stop more than once while waiting for an i for install input.
Type sudo apt install curl and accept all prompts. There will be additional packages. This tool lets you run scripts in terminal windows.
Type carefully or copy-paste this text exactly: curl -fsSL https://ollama.com/install.sh | sh. The funny bar character is called a "pipe" and is usually shift- \ backslash on your keyboard.
A ten to twenty minute download will add a gigabyte of data. Watch for Install complete.
Another download so that the ollama application can work with the deepseek database. At the time of this writing ollama run deepseek-r1:1.5b does work. Be careful of dots, dashes, slashes, caps and miniscules.
Progress bars for several minutes. More than another gigabyte. We return to a >>> prompt from ollama. I suggest typing /bye and exiting to start fresh.
Open ollama again. It should not need to update and we can ask a question.
If that is it then I am underwhelmed. This technology may be over-hyped.
/? helpThere's a help feature /?. /bye returns us to the shell outside of the application.
The first control-c will return to ollama prompt. Second control-c exits application.
We've enabled a server by installing ollama. This command tells us if it's listening.
This is a release of Ubuntu managed by Nvidia and not UbuntuCorp. apt package manager works with their repositories and you can install Synaptic. It can run movie and music players, function as a wordprocessor with LibreWrite.
Install hardinfo system profiler and get a closer look at your machine.
If you see new software releases on the Nvidia website please comment.
Comments