Nvidia Corp. is moving to lower the entry barrier to developing artificial intelligence models.

The chip maker, which sells graphics processing units commonly used for running AI applications, today updated its Nvidia GPU Cloud to support everyday desktops. Machines running the company’s consumer-oriented Titan accelerators can now take advantage of deep learning capabilities that were previously available only on enterprise platforms.

“It’s really just blowing open access to the software,” Jim McHugh, Nvidia’s vice president and general manager of enterprise systems, said in an interview. “This is about bringing it to the masses.”

GPU Cloud provides a set of AI development kits optimized for Nvidia chips. The lineup includes prepackaged versions of popular open-source deep learning engines such as Google LLC’s TensorFlow framework, the Caffe2 platform created by Facebook Inc. and several others. Each tool comes bundled with the various software components that Nvidia offers to help developers make the most out of its silicon for their AI projects.

These kits are provided alongside a similar lineup of packages for performing data visualization. The ability to explore information in a graphical form is essential for engineering and scientific use cases, which are a big driver of GPU sales.

Nvidia said the preconfigured nature of its packages can cut the considerable amount of manual work usually involved in configuring an AI development environment. The bundles are delivered as software containers, which enables companies to set them up on many different types of infrastructure.

Nvidia said that has the added benefit of providing portability. As a result, a developer can prototype an AI model on their desktop and easily push the code to their company’s backend environment when more computing resources are needed. The feature avoids some of the challenges normally associated with moving something as complex as a deep learning model from one type of infrastructure to another.

The fact that Nvidia’s GPU Cloud can now work with a regular Titan-powered desktop should enable more developers to take advantage of its features. The chip maker has a vested interest in fostering the creation of AI models, since a good portion of the ones running today are powered by its chips.

“Most developers start on their own desktop,” McHugh said. “Now those in research or academia or even weekend researchers can do it.”

The new desktop compatibility is joined by the addition of support for two new deep learning frameworks. The first is the PaddlePaddle engine that Chinese search giant Baidu Ltd. released last year, which allows developers to implement certain models with a lot less code than some alternatives. The other is the 1.0 release of MXNet, the AI framework backed by Amazon.com Inc.’s cloud division.

With reporting from Robert Hof

Image: Nvidia