sciwork 2023

uTensor: Deep Learning Inference Engine Born for TinyML
2023-12-10, 13:30–14:00 (Asia/Taipei), NYCU

This talk mainly introduces the origin and evolution of uTensor, covering the related API design concepts and embedded system issues, including memory management and heterogeneous storage, etc.


As a new frontier in machine learning, TinyML combines embedded systems, scientific computing, and deep learning. At the time, there was a need for a tool that was easy to use, debug, expand, and allowed embedded system engineers and data scientists, especially deep learning experts, to collaborate seamlessly. Thus, uTensor was born.

In this talk, in addition to providing a basic introduction to uTensor, we will also share the history and evolution of uTensor, as well as discuss related topics on integrating with embedded systems, including memory management and heterogeneous storage issues.

Slide

Ask questions at slido


Prior Knowledge Expected?

No, previous knowledge expected

Language

Mandarin talk w. English slides

YAMLE, Yet Another Machine Learning Engineer.
That's me, deal with it 😉