As part of IBM’s continuing effort to expand its AI portfolio, IBM is adding Anaconda, an open source analytics platform, to its Power-based cognitive computing offerings, as well as integrating it into its own PowerAI software distribution.
Anaconda is an open source software suite that encapsulates over 720 data science packages, including those that support data visualization, AI/machine learning, and other types of advanced analytics. Developed by Continuum Analytics, Anaconda is touted as “enterprise-ready Python for big data.” Although Python is traditionally an interpreted computer language, the version used by Anaconda is a high-performance distribution compiled with Intel’s Math Kernel Library. As a result, it’s well-suited to numerically-intensive applications such as those involved in machine learning, predictive analytics, and scientific computing.
Anaconda can be had for free from Continuum Analytics’ website, with different versions availabel for Windows, macOS, and Linux. According to the company, the software has been downloaded more than 16 million times.
IBM is adding Anaconda to bolster its cognitive computing ecosystem, specifically its Power systems that offer NVIDIA’s Tesla GPU accelerators. The current rendition of this platform is based on the Power Systems S822LC for High Performance Computing server, also known as the “Minsky” platform. It puts two Power8 CPUs with up to four NVIDIA Tesla P100 GPUs in the same box. Currently the Power8 is the only CPU that offers a high-bandwidth interface to these GPUs, via NVIDIA’s NVLink technology. The interface provides much faster communication between the CPU and GPU than a typical PCIe Gen3 connection – on the order of 2.5 times as fast.
The Anaconda integration will enable users of this open source software to build applications for the deep learning-optimized Minsky hardware. In particular, it will enable IBM to leverage its PowerAI software toolkit, a collection of deep learning frameworks like Caffe, TensorFlow, and Theano. The Anaconda addition will offer prospective IBM customers the ability to build these types of applications that can run on their hardware.
In conjunction with the Anaconda news, IBM also announced that the OpenPOWER Foundation has formed a Machine Learning Work Group that will focus on expanding such development within the OpenPOWER community. The new group will include Google, Mellanox, and Cineca, among others. Machine learning will also be the focus topic at the upcoming OpenPOWER Foundation Developer Congress. The event will take place from May 22-25 and be held at the Palace Hotel in San Francisco.