30 Golden Rules of Deep Learning Performance


Duration: 50 mins
Siddha Ganju
Self-driving Architect, NVIDIA

“Watching paint dry is faster than training my deep learning model.”

“If only I had ten more GPUs, I could train my model in time.” “I want to run my model on a cheap smartphone, but it’s probably too heavy and slow.”

If this sounds like you, then you might like this talk.

Exploring the landscape of training and inference, we cover a myriad of tricks that step-by-step improve the efficiency of most deep learning pipelines, reduce wasted hardware cycles, and make them cost-effective. We identify and fix inefficiencies across different parts of the pipeline, including data preparation, reading and augmentation, training, and inference.

With a data-driven approach and easy-to-replicate TensorFlow examples, finely tune the knobs of your deep learning pipeline to get the best out of your hardware.

You may also be interested in

180 mins
Building Reactive Applications

Reactive Programming is receiving quite a bit of attention and for good reasons. It's a nice logic next step from...

50 mins
Maximizing Performance with GraalVM

In this session we will discuss the best practices for Java code and compiler configurations to maximize performance with GraalVM...

50 mins
A State of Java Elasticity

Java is finally elastic! OpenJDK improvements and new features in Garbage Collection technology resulted in enhancing Java vertical scaling and...

50 mins
Jakarta EE 9 and Beyond

Jakarta EE 9 is out! This is the second release of Jakarta EE since the transition from the Java Community...

50 mins
Exploring Collectors: One of the Most Powerful Utility Class in the JDK

One of the most intriguing classes in the JDK is the Collectors utility class, with a collection of some highly...

50 mins
Beginning Java 9 Modules

Java 9 modules present a radical shift away from how we thing about modularizing our applications. However, as we have...