Accelerating Deep Learning Models on Xilinx 7nm Versal Card

preview_player
Показать описание
Xilinx 7nm Versal developer card for AI, VCK5000, is now available with one of the highest performance/$ scores on the latest MLPerf 1.0 benchmarks. Join this webinar to learn how to directly take your TensorFlow / Pytorch models and accelerate on the VCK5000 card. This session will cover the introduction of three overlays on CNN and NLP application with a demo showing how to get deep learning models running on VCK5000.

00:00 Agenda
00:56 VCK5000 Introduction
02:33 Why so many IP Overlays
04:10 DPU for CNN
13:08 MLPerf v1.0
15:42 DPU for BERT
21:45 Demo of DPU for CNN
Рекомендации по теме