GTC 2020: How NVIDIA’s Deep Learning Training Examples have State-of-the-Art Accuracy and Performance

GTC 2020 S21765
Presenters: Pablo Ribalta,NVIDIA
Abstract
Convert ideas into fully working solutions with NVIDIA Deep Learning examples. Have you ever scraped the net for a model implementation and ultimately rewritten your own because none would work as you wanted? Get as fast as possible to a working baseline by pulling one of our many reference implementations of the most popular models. They come with a step-by-step guide, can be deployed directly in our latest NVIDIA container, and are optimized to deliver state-of-the-art accuracy with full reproducibility across different GPU setups. Curious about mixed precision or multi-GPU? We have those, too! Implementations are available in Tensorflow and Pytorch, with interactive examples consisting of Jupyter Notebooks.

Watch this session
Join in the conversation below.