It seems as if finding the right Amazon Machine Image (AMI) ID for the workload you’d like to deploy can sometimes be a bit of a challenge. Each combination of region and AMI produces a unique ID, so you have to look up the AMI for the particular region where you’re going to deploy the workload. This in and of itself wouldn’t be so bad, but then you have to wade through multiple versions of the same AMI in each region. Fortunately, if you’re using CoreOS Container Linux on AWS, there’s an easy way to find the right AMI ID. Here’s how it works.
CoreOS publishes a JSON feed of the latest AMI for each of their channels (stable, beta, and alpha). You can find links to these JSON feeds on this page. This is powerful for 2 reasons:
Because it’s available via HTTP, you can use curl to retrieve it anytime you need it.
Because it’s in JSON, you can use jq (see my post on jq for more information) to easily parse it to find the information you need. (Not super comfortable with JSON? Check out my introductory post.)
Putting these two reasons together, you end up Continue reading
It is one thing to scale a neural network on a single GPU or even a single system with four or eight GPUs. But it is another thing entirely to push it across thousands of nodes. Most centers doing deep learning have relatively small GPU clusters for training and certainly nothing on the order of the Titan supercomputer at Oak Ridge National Laboratory.
The emphasis on machine learning scalability has often been focused on node counts in the past for single-model runs. This is useful for some applications, but as neural networks become more integrated into existing workflows, including those …
Scaling Deep Learning on an 18,000 GPU Supercomputer was written by Nicole Hemsoth at The Next Platform.
The post Worth Reading: Odds are in Favor of Quantum Encryption appeared first on 'net work.