Running Generative AI on a $1 Raspberry Pi Chip
Generative AI models tend to have enormous hardware requirements. Even running a relatively small image generation model locally generally requires a decent computer with a good amount of RAM. And if you don’t want to wait all day for an image, you will probably want a powerful GPU as well. A $1 microcontroller with 520 KB of RAM is about the last place you would expect to find one of these models running.
But a developer named Tim has been interested in making that happen for quite some time. A while back, he began experimenting with the idea of implementing a generative AI image model on a microcontroller. He got as far as prototyping a diffusion model and conditional variational autoencoder (VAE), but never actually deployed the system to a microcontroller. He recently returned to the idea and this time succeeded.
The hardware used in the project (📷: Tim)
His ...
Copyright of this story solely belongs to www.hackster.io. To see the full text click HERE
