This event has passed
In this meetup, we're going to feature two practical approaches to using GPUs to accelerate training AI.
First, Nick will take a workload that trains an image classification neural net and show how you can leverage Nvidia GPUs to see impressive speed improvements without any substantial code changes. We'll go over the differences in model training speed on a local machine or cluster as compared to training the same model in a more distributed environment using cloud resources that can be spooled up or torn down as needed. You'll see the use of popular libraries such as Tensorflow and Keras as well as a popular image classification model ResNet.
Everyone knows that GPUs almost always process data faster than CPUs, but if you're used to the tried and true python libraries based on CPU processing, like pandas and sklearn, you might be hesitant to refactor old code or learn a new library for your data science work to take advantage of GPU performance.
In our second presentation, Jake will demonstrate in an end-to-end ML project how RAPIDS.ai libraries like cudf and cuml have removed the complexities of working with GPUs and how you can take advantage of these libraries.
For a preview of what we'll be covering, we've got the following resources:
Cloudera Users Page:
Join Jake Bengtson, Nicolas Pelaez and Michael Ridley, all of Cloudera, to see how we've linked all these concepts together and hopefully inspire similar solutions of your own!
This is still a tricky time for public gatherings, but Future of Data is committed to providing great tech content & facilitating discussions in the "Big Data" and machine learning space. Our group in Northern Virginia is holding this event; in order to do our part to fight the spread of COVID-19's delta variant, this will be an exclusively online event with an originating Time Zone of EDT (the event time displayed on this page will reflect the equivalent local time). We thought it might be of interest to our wider membership (you are welcome to sign up for it here).
The URL for accessing the "live stream" will be provided to pre-registrants here, on this page, no later than 48 hours prior to the event.