Dynamically loading expert weights on the Neural Engine

Machine Learning & AI Core AI
atilaa OP
Created 4h
Replies 0
Boosts 1
Views 109
Participants 1
This post is from the WWDC26 Core AI Q&A.

AFM3 Core Advanced dynamically loads experts into DRAM* for Neural Engine execution. Is this capability accessible to third party via Core AI or Core ML APIs in iOS 27?

Reference: https://machinelearning.apple.com/research/introducing-third-generation-of-apple-foundation-models

Replies  0
Boosts  1
Views  109
Participants  1
Dynamically loading expert weights on the Neural Engine
First post date Last post date
 
 
Q