Acumos serves up its first software release to fuel AI

The Acumos AI Project debuts its first software release, which is called Athena. (Pixabay)

Open source community Acumos AI on Wednesday announced the availability of its first software release, which is called "Athena."

The Acumos AI Project is in under the Linux Foundation's LF Deep Learning Foundation umbrella, which was designed to spur innovation across artificial intelligence, machine learning and deep learning not just in the telecom industry, but across other industries as well.

Amdocs, AT&T, B.Yond, Baidu, Huawei, Nokia, Tech Mahindra, Tencent, Univa and ZTE were the founding members of LF Deep Learning back in March, followed by new members Ciena, DiDi, Intel, Orange and Red Hat in August.

While the current membership is impressive, several key service providers are missing from Deep Learning's roster including Verizon, BT, CenturyLink, Deutsche Telekom and Telefónica.

AT&T and Tech Mahindra announced in October of last year that they had partnered to create the Acumos platform with the goal of putting it in to open source. Acumos was created to simplify the development of AI while also providing a marketplace for accessing, using and enhancing AI apps.

While Acumos was the first announced project for Deep Learning, it has since added Angel and  EDL, which were developed by Baidu and Tencent.

Some of those initial goals are reflected in the Athena release, which includes one-click deployment of the platform by using Docker or Kubernetes. Athena also provides the ability to deploy models in public or private cloud infrastructure, or in a Kubernetes environment on users' own hardware, such as servers, and virtual machines.

There's also a design studio element in Athena, which includes a graphical interface for chaining together multiple models, data translation tools, filters and output adapters onto an end-to-end platform.

On the marketplace front, Athena has a user portal that lets users personalize marketplace views by theme, and data on model authorship as well the ability for sharing models privately or publicly. It also provides upgrades to user experiences.

"The Acumos AI Marketplace, (is) a catalog of community-contributed AI models with common APIs that can be shared securely across multiple systems, remains open and accessible to anyone who wishes to download or contribute models and applications," said LF Deep Learning Outreach Committee Chair Jamil Chawki in a prepared statement.

Athena also features a security token that allows onboarding of models from an external toolkit directly into an Acumos repository. There's a feature for the decoupling of microservices from the model onboarding process, which enables re-purposing of the models for different hardware and environments.

“The Acumos Athena release represents a significant step forward in making AI models more accessible for builders of AI applications and models along with users and trainers of those models and applications,” said Scott Nicholas, senior director of strategic planning at The Linux Foundation, in a prepared statement. “This furthers the goal of LF Deep Learning and the Acumos project of accelerating overall AI innovation.”

The developer community for Acumos AI is working on the next release, which will be available in mid-2019. The second software release will include model training as well as data extraction pipelines to make models more flexible. It will also have  updates to assist closed-source model developers, including secure and reliable licensing components to provide execution control and performance feedback across the community.

The importance of Acumos

During a previous interview with FierceTelecom, AT&T Labs' Mazin Gilbert, vice president of advanced technology and systems and Governing Board Chair of LF Deep Learning, said the Deep Learning Foundation reaches beyond telecom into other sectors such as healthcare, security, real estate, and enterprise.

RELATED: AT&T's Mazin Gilbert: Technologies are great, but convergence is the most exciting part

Mazin said the Deep Learning Foundation umbrella of projects would align with the Linux Foundation Networking Fund (LFN) on projects such as ONAP.

"So, Acumos and Deep Learning are not about telecom. It's broad," Gilbert said. "You're going to see some announcements made soon by some of our other companies in areas that have nothing to do with telecommunications. It was purposely done that way.

"From an AT&T perspective, we're going to create that link, between AI and 5G, 5G and Linux Foundation Networking ONAP. Between LFN and ONAP and Akraino. I call it the diamond."

Tungsten Fabric launches Performance Lab

In other Linux Foundation-related news, LFN announced a new Tungsten Fabric Performance Lab. Launched jointly by Intel, Juniper Networks, Lenovo and SDNLAB, the Performance Lab aims to forge an open source test and verification environment for Tungsten Fabric and other Software Defined Networks (SDN) projects.

Tungsten Fabric is an open source, scalable, multi-cloud, multistack networking platform that provides a single point of control and analytics for networking and security. Tungsten Fabric, which was developed by Juniper Networks before it was put into open source, is integrated with private cloud stacks including Kubernetes, VMware and OpenStack. It also supports hybrid deployments with public clouds including AWS and Google Compute Engine.