It's Apple's API for hardware-accelerated machine learning. Their new devices have specialized ML hardware so it's actually feasible to run on mobile devices. Probably what they also use for Apple Intelligence. https://developer.apple.com/documentation/coreml
12 Comments
JOMusic@lemmy.ml · 5 pts · 1y
GitHub here: https://github.com/allenai/OLMoE.swift
Ai2 make as much of their training data available as possible: https://huggingface.co/datasets/allenai/OLMoE-mix-0924
kekmacska@lemmy.zip · 3 pts · 1y
Wanna see how you git clone and compile it on ios
jeena@piefed.jeena.net · 3 pts · 1y
I wonder how bit it is, normally they call them something with 3B or 7b
oktoberpaard@feddit.nl · 1 pts · 1y
It’s 7 billion parameters big.
jeena@piefed.jeena.net · 1 pts · 1y
Ah, nice, not bad for a iPhone model.
foremanguy92_@lemmy.ml · 3 pts · 1y
Why only iPhone?
2xsaiko@discuss.tchncs.de · 6 pts · 1y
It probably uses Apple’s CoreML. Not sure if Android has an equivalent API.
fubarx@lemmy.ml · 3 pts · 1y
https://developers.google.com/ml-kit/
foremanguy92_@lemmy.ml · 1 pts · 1y
Explain it pls
2xsaiko@discuss.tchncs.de · 3 pts · 1y
It's Apple's API for hardware-accelerated machine learning. Their new devices have specialized ML hardware so it's actually feasible to run on mobile devices. Probably what they also use for Apple Intelligence. https://developer.apple.com/documentation/coreml
foremanguy92_@lemmy.ml · 1 pts · 1y
Seems like a good project to implement on the green side (Or pinguin side)
JOMusic@lemmy.ml · 2 pts · 1y
according to the linked article, it sounds like it is for performance reasons.