Follow Datanami:
January 26, 2017

IBM: We’re the Red Hat of Deep Learning

IBM today took the wraps off a new release of PowerAI, the prepackaged bundle of deep learning frameworks that debuted last fall. With the addition of Google’s TensorFlow framework, the company says its AI business model is starting to resemble the Linux distributor Red Hat.

“In a sense, PowerAI makes IBM the Red Hat of deep learning,” Sumit Gupta, the vice president of IBM‘s High Performance Computing & Data Analytics business, told Datanami. “You don’t go to Linux.org to download Linux. You go to Red Hat. In the same way, instead of going to TensorFlow or Caffe or other websites [for deep learning frameworks], they want an enterprise-level distribution.”

The first release of PowerAI included “optimized” versions of Caffe-bvlc, Caffe-ibm, Caffe-nv, DIGITS, Torch, and Theano. With this release, the software includes TensorFlow 0.12, as well as Chainer, a deep learning framework that’s very popular in Japan.

Gupta says PowerAI—which is free and distributed as a binary for Ubuntu Linux (sorry, Red Hat)—is resonating with IBM clients, particularly those who have bought the “Minsky” Power8 servers to run deep learning and machine learning workloads on.

“Enterprise customers prefer not to go to an open source website and download software and build it. All that stuff is not what a data scientist wants to do,” he says. “What we’ve done is essentially created an easy-to-deploy solution. You want to do deep learning? Buy a Minsky, download the free PowerAI software, and boom–you’re ready to go.”

IBM also announced results of benchmark tests that show Minsky (Power S822LC for High Performance Computing) runs TensorFlow models 30% faster than an X86-based system equipped with PCI-based GPUs. The Minsky server, of course, combine Power8 CPUs with NVIDIA Pascal P100 GPUs via fast NVLink connections. The NVLink interface is not yet available for X86-based CPU-to-GPU connections.

“Thirty percent is huge because a lot of these deep learning training runs take days, and in some cases more than a week to run,” Gupta says. “When you’re talking about 30%, you’re cutting off days in training time.”

While PowerAI itself is free, users must develop their own models and applications. IBM is committing to keeping up with the rapid pace of innovation in the frameworks, and to put its stamp of approval on those pieces of software.

IBM today also announced a new technical services offering to develop PowerAI-based deep learning and machine learning models and applications for clients through IBM Global Business Solutions. This service will likely interest those companies that lack the sophisticated data science expertise to develop their own models and algorithms, but that want to begin benefiting from big data analytics just the same.

The new IBM Power S822LC for High Performance Computing, codenamed “Minsky”

While IBM is not sharing the names of companies that are using PowerAI, Gupta assured us that it’s being very well received by the market, and that companies in various industries are using PowerAI for a variety of use cases.

Many of the PowerAI use cases involve computer vision, which refers to computers identifying objects captured with a camera. Among the applications here are drone-based industrial inspection, surveillance, and self-driving cars.

“We’re seeing retail customers try to build recommendation engines,” Gupta says. “We have finance customers trying to do fraud detection.  There’s a pretty broad set of use cases out there.”

You can download PowerAI at www.ibm.com/us-en/marketplace/deep-learning-platform.

Related Items:

IBM Debuts Power8 Chip with NVLink and 3 New Systems

All Eyes On AI in 2017

IBM ‘DataWorks’ Leverages Watson, Spark

Datanami