5 ms·
Tiny-diffusion: A minimal implementation of probabilistic diffusion models
- aswanson 1y agoThis is where I see edge ai going. Minimal implementations running on embedded systems everywhere.
- endymion-light 1y agoFull agree - this is where AI is finding it's main useful niche. Massive models are fun to use but specialization within embedded systems is the clincher.
- sitkack 1y agoHaving AI running in every edge location poses an existential threat. They will provide the peripheral sensors for a larger intelligence.
- stereobeings 1y ago[dead]
- barbarr 1y agoVery cool showing the outputs of hyperparameter search. Hopefully this gets someone out of a tuning tarpit!
- aestuans 1y agoThis is very cool. I implemented a slightly more complicated version of this with class guidance a while ago: https://github.com/aestuans/mnist-diffusion https://github.com/aestuans/mnist-diffusion