HPE says the new cloud offering will be 100% carbon neutral and will run on the Cray supercomputer.
GreenLake, a new supercomputing cloud service for large language models, will be available in the United States in late 2023 or early 2024, Hewlett Packard Enterprise announced Tuesday at HPE Discover. GreenLake for LLM enables enterprises to train, tune, and deploy large-scale artificial intelligence private to their individual businesses.
GreenLake for LLM will be available to customers in Europe following the U.S. release, and is expected to be released in early 2024.
Jump to:
HPE partners with AI software startup Aleph Alpha
“AI is at an inflection point and at HPE we are seeing demand from a variety of customers who are beginning to leverage generative AI,” said executive vice president and general manager of the HPC & AI Business Group and Hewlett-Packard Research Institute. Justin Hotard said in a virtual presentation. .
GreenLake for LLM runs on AI-native architecture across hundreds or thousands of CPUs or GPUs, depending on your workload. This flexibility within a single AI-native architecture product makes it more efficient than general-purpose cloud options that run multiple workloads in parallel, he said. GreenLake for LLM was created in partnership with German AI startup Aleph Alpha, which offers a pre-trained LLM called Luminous. Luminous LLM works in English, French, German, Italian and Spanish and can make predictions using text and images.
The collaboration was two-way, with Aleph Alpha initially using HPE infrastructure to train Luminous.
“Using HPE supercomputers and AI software, we were able to train Luminous efficiently and quickly,” said Jonas Andrulis, founder and CEO of Aleph Alpha, in a press release. . “We are proud to be the launch partner of HPE GreenLake for Large Language Models. We have expanded our collaboration with HPE to extend Luminous to the cloud and offer it as a service to end users, We look forward to facilitating new applications in business and research initiatives.”
The first release will include a set of open source and proprietary models for retraining or fine tuning. In the future, HPE will offer specialized AI for tasks related to climate modeling, healthcare, finance, manufacturing and transportation.
For now, GreenLake for LLM will be part of HPE’s overall AI software stack (Figure A), which includes Luminous models, machine learning development, data management and development programs, and the Cray programming environment.
Figure A
HPE’s Cray XD supercomputer delivers enterprise AI performance
GreenLake for LLM runs on HPE’s Cray XD supercomputers and NVIDIA H100 GPUs. Supercomputers and his HPE Cray programming environment allow developers to work on data analysis, natural language tasks, and other high-performance computing and AI applications without running their own hardware. . This can be costly and require expertise specific to supercomputing.
During the presentation, Hotard pointed out that AI production in large enterprises requires enormous performance resources, skilled people, security and trust.
look: NVIDIA offers AI tenancy on DGX supercomputers.
Get more power from renewable energy
By using colocation facilities, HPE aims to power supercomputing with 100% renewable energy. HPE is working with North American compute center specialist QScale on a design built specifically for this purpose.
“Our goal with all our cloud deployments is to provide our customers with a 100% carbon-neutral service,” said Hotard. “One of the advantages of liquid cooling, for him, is that wastewater or heated water can actually be taken out and reused. I am making use of it.”
Alternatives to HPE GreenLake for LLM
Other cloud-based services for running LLM include NVIDIA’s NeMo (currently in early access), Amazon Bedrock, and Oracle Cloud Infrastructure.
During his presentation, Hotard said GreenLake for HPE is not a replacement for, but a complement to, large cloud services such as AWS and Google Cloud Platform.
“We can and will integrate with the public cloud. We see this as a free service. We don’t see it as a competitor.” He said.
