D-Matrix lands $44 million to build AI-specific chipsets

HPE announces integrated appliance for machine learning development

Posted on

Hewlett Packard Enterprise Co. today is unveiling some of the fruits of last year’s acquisition of Determined AI Inc. with the launch of a platform for building and training machine learning models at scale.

The HPE Machine Learning Development System combines the HPE Machine Learning Development Environment with computing, accelerators and onboard networking to accelerate model development significantly, the company said.

It’s intended to address what is an often complex, multistep process that’s involved in purchasing and installing massively parallel processors incorporate specialized compute, storage, interconnect and accelerators. The packaged offering, which is available now, allows organizations to begin building and training machine learning models immediately.

“Training deep learning models is not only complex and time-consuming but resource-intensive,” said Justin Hotard, general manager of high-performance computing, mission-critical solutions and labs at HPE. “Many of the engineers spend their time managing infrastructure rather than focusing on optimizing their models. This means they can focus on business outcomes instead of technology requirements.”

The system will be offered as a single package based on the HPE Apollo 6500 Gen10 system and starting with eight Nvidia Corp. A100 80-gigabyte graphic processing units. The management stack uses HPE ProLiant DL325 servers and a 1Gb Ethernet Aruba CX 6300 switch.

Networking and storage are provided by Nvidia Quantum InfiniBand and monitoring and management by HPE Performance Cluster Management.

“What generally exists in the market today are rigid solutions that become quite costly at scale,” Hotard said. “That means greater complexity and longer time to insight for customers.” HPE’s objective is to give customers “great flexibility in where they deploy models and the infrastructure they deploy those models on.”

HPE is also building on its artificial intelligence presence with the launch of HPE Swarm Learning, a privacy-preserving, decentralized machine learning framework for edge or distributed computing. The framework provides customers software containers that can be integrated with AI models using the HPE swarm API. It enables organizations to share learnings from AI models with other organizations without sharing any actual data.

Most AI model training relies on centralized, merged datasets, which is both inefficient and costly because of the need to move large volumes of data. In regulated industries, such training can also be subject to data privacy and ownership rules that limit external data sharing and movement. The result is lower-quality AI models, HPE said.

HPE Swarm Learning enables organizations to use distributed data at its source, which increases the dataset size for training while preserving data governance and privacy rules. Blockchain technology is used to secure onboard members, dynamically elect a leader and merge model parameters.

Additionally, HPE announced it’s building on its collaboration with Qualcomm Technologies Inc. to deliver advanced inferencing offerings in support of heterogeneous system architectures for AI inferencing large scale. HPE will offer the ruggedized Edgeline EL8000 Converged Edge systems along with Qualcomm Cloud AI 100 accelerator for inferencing and at the edge. The offering will be generally available in August.

Photo: Pixabay

Show your support for our mission by joining our Cube Club and Cube Event Community of experts. Join the community that includes Amazon Web Services and Amazon.com CEO Andy Jassy, Dell Technologies founder and CEO Michael Dell, Intel CEO Pat Gelsinger and many more luminaries and experts.

Source link

Leave a Reply

Your email address will not be published. Required fields are marked *