Boosting ML Networks on Specific Hardware Platforms with Apache TVM - Example: Qualcomm Adreno GPU

Explore the power of Apache TVM, a tensor compiler for optimizing machine learning networks, with a focus on boosting performance on Qualcomm Adreno™ GPUs. Learn how to leverage hardware-specific features to significantly improve model execution speed. Discover the journey of optimizing Apache TVM for Adreno GPUs, from initial research in 2021 to substantial performance enhancements achieved in 2022. Gain insights into running models efficiently on target platforms and implementing best practices for optimal performance. This 31-minute presentation, delivered by Egor Churaev, a Senior Software Engineer at Deelvin Solutions and Apache TVM committer, offers valuable knowledge for developers and researchers looking to maximize ML network performance on specific hardware platforms.