Earlier this year, Microsoft released an open source artificial intelligence toolkit on GitHub called CNTK. The toolkit has been originally created by Microsoft researchers to leverage deep learning and provide speech and image recognition capabilities more efficiently than other popular computational toolkits.
Today, Microsoft has announced the release of a new version of the toolkit, which is now called Microsoft Cognitive Toolkit. The project is no longer a research tool as the company is now using it in a production setting to improve its various AI-based products such as its digital assistant Cortana. The latest version of the toolkit brings some performance improvements plus new capabilities such as support for Python or C++ programming languages.
Chris Basoglu, a partner engineering manager at Microsoft who worked on the toolkit explained in the blog post that “one key reason to use Microsoft Cognitive Toolkit is its ability to scale efficiently across multiple GPUs and multiple machines on massive data sets.” For now, the toolkit is optimized for NVIDIA GPUs but it can also run on Microsoft’s Azure N-Series GPU offering which is still in preview.