tinyML Talks local Webcast – recorded February 18, 2021
“An Introduction to Optimizing ML Models with TVMC”
Chris Hoge – OctoML
Currently, deep learning systems rely on vendor-provided kernel libraries or various search strategies to get performant tensor programs. This talk will give an introduction to TVMC, a front-end to the Apache TVM deep learning compiler stack that can build optimized models for a variety of different platforms. It will cover how TVM works at a high level, then give a practical introduction on how to use TVMC in practice.