Rapid growth in the deep learning software ecosystem has made deploying trained models into production harder than it should be. In the search for performance, users need to consider hardware specific vendor specific libraries, inference engines, tuning parameters, and so on, then manage the often incomplete operator coverage, performance cliffs, and portability pitfalls that you face when evaluating these software stacks. In this talk we will discuss Apache TVM, an open-source, end-to-end deep learning compiler stack that directly addresses the challenge of deploying highly optimized models to a variety of current and future hardware. The motivation, architecture, and performance of models generated by TVM will be covered, along with information on how to use TVM and become involved with its growing community.

Note, this virtual event will be held using zoom. The link will be shared the day before the event.

ZOOM Link: https://zoom.us/j/92356581624?pwd=NHhjUUhDbHl1Q0V5TVRMQzdPY0gzdz09

Agenda

5:00 pm: Welcome & Housekeeping

5:05 pm: "TVM: An End-to-End Deep Learning Compiler Stack" (Chris Hoge, Developer Advocate and Jason Knight, CPO @ OctoML.ai)

5:45 pm: Q&A and Networking

Hosted By: Rsqrd AI