NVAITC AI Webinar Series on Deep Learning
Weekly, Thu 17 Sep – Thu 1 Oct 2020, 09:00–10:00
The goal of this webinar series is to explore the fundamentals of deep learning by building and training neural networks, optimizing data loading and performance through mixed- precision arithmetic and parallelization across several GPUs. Furthermore, we demonstrate how to deploy your trained model in production for inference. You will learn how to design, train, optimize, profile, and deploy a deep neural network using NVIDIA technology.
Target audience
Master’s students, PhD students, postdocs, researchers working in the field artificial intelligence
How to follow the webinars
The webinar series will be presented live using Zoom web conferencing system. Please install the Zoom client for best experience but you can also join via a standard web browser.
Meeting ID: 494 152 0325
Passcode: Finland
The seminars are recorded and will be made available afterwards. Please regularly check this online Drive for the webinar slides and recordings.
Agenda
Thursday 17 September 2020, 9.00–10.00 – Getting started with DL
Session 0: Linear Regression in Pytorch (20’ + 10’ q/a) - Christian Hundt
Session 1: Convolutional Neural Networks (20’ + 10’ q/a) - Christian Hundt
Thursday 24 September 2020, 9.00–10.30 – Fast and scalable training
Session 2: Efficient Data Loading using DALI (20’ + 10’ q/a) - Giuseppe Fiameni
Session 3: Mixed Precision Training using Apex (20’ + 10’ q/a) - Paul Graham
Session 4: Multi-GPU Training using Horovod (20’ + 10’ q/a) – Gunter Roeth
Thursday 1 October 2020, 9.00–10.00 – Model deployment and profiling
Session 5: Deploying Models with TensorRT (20’ + 10’ q/a) - Niki Loppi
Session 6: Profiling with NVTX (20’ + 10’ q/a) - Giuseppe Fiameni