AI Model Efficiency Toolkit (AIMET) - Lecture 25

Explore the world of AI model efficiency in this guest lecture from Qualcomm AI Research. Dive into the challenges of deploying neural networks on mobile and IoT devices, and discover cutting-edge solutions for efficient machine learning. Learn about Qualcomm's core technologies, adaptive rounding, autocon, and transformer quantization. Gain insights into image model zoos, conditional compute, and the Qualcomm Snapdragon 8 Gen 2. Witness a live demo showcasing the practical applications of these efficiency techniques. Perfect for those interested in TinyML, efficient deep learning computing, and the future of AI on resource-constrained devices.

Introduction
Challenges
Qualcomm AI Research
Qualcomm Core Technologies
Layers of Interest
Inference
Features
Adaptive Rounding
Autocon
Training
Image Model Zoo
Transformer Quantization
Snapdragon Gen 2
GitHub
Qualcomm Snapdragon 8
Conditional Compute
Demo