The Raspberry Pi has been out for a few months now and we’ve had a chance to get a close look at what it’s capable of. Now maker and developer Data Slayer has delved into using a Pi 5 to run advanced AI large language models (LLMs) in his latest video. In it, we get an impressive look at the Pi 5’s ability to utilize local AI models.
Data Slayer is using a tool called Olama to manage and switch between the LLMs. This is done via command line and doesn’t require any access to the internet as everything is stored on the Pi 5. In the video, Data Slayer gives us a close look at what a few different models can do when operating on the Pi 5.
The first LLM he demonstrated is called LLaVa and is intended for analyzing pictures. To put it to the test, Data Slayer snapped a selfie and parsed it through the Pi to see what LLaVa would detect. It correctly identified the subject of the picture noting that it was of a man wearing a hoodie in a building located somewhere in an urban environment.
The next example was a model known as Llama 2 which handled some amusing text-based inquiries. Right off the bat, he prompted it for a spicy mayonnaise recipe and then proceeded to test it with a little bit of historical trivia. To finish off the experiment, he asked it to generate a few basic Linux commands.
This is a really neat example of what you can do with the Raspberry Pi 5 without any extra hardware. That said, you can definitely enhance the experience by adding storage or an external TPU to improve the performance. In the video put together by Data Slayer, he’s using an 8GB Raspberry Pi 5 with a 256GB microSD card.
If you want to get a closer look at this Raspberry Pi project in action or just want to know more about how it all works, check out the original video shared to YouTube by Data Slayer and be sure to follow him for future creations.
