Getting More from Your GPU – Tutorial by Yuval Mazor

NOTE:

• Link to slides in presentation – Getting More from Your GPU
• Due to a technical problem there is no recording of this, however, here is a link to a youtube video by Yuval on A Practical Guide for Reducing DNN Training Time

————————————————————————————————–

This is a computing Tutorial hosted and organized by the DSI open for DSI members and their graduate students.

Please register to event via following link: https://forms.gle/XUkyGacZfMCKvta18

When: May 4, 2023. 15:00-16:30
Where: building 604, room 62

Title:  Getting More from Your GPU

Abstract:

GPUs are complex creatures that are easy to use but not necessarily easy to use right.  In this talk we’ll be taking a look at some issues having to do with using GPUs properly:

1.  What do the letters and numbers mean?  (A100, V100, A10, 3090, etc.)  When should I use what kind?
2.  How do I correctly run multiple processes on a single GPU?
3.  What do I need to understand when running models on multiple GPUs?
4.  Best practices for your model

Note: Session will be recoded but not broadcasted live.

Short bio:

Yuval Mazor is an experienced data scientist with over 20 years of industry experience as a developer, architect and technical leader in algorithms, software design and implementation and machine/deep learning.  He has worked in diverse environments, ranging from 2-person startups to large multi-nationals.  Yuval is passionate about technology and its implications to daily life – as such he speaks often in user groups and conferences.  Yuval has a Bachelor’s degree in Computer Science and has done graduate work in Computational Linguistics and Natural Language Processing at Tel Aviv University.