Generative AI
Generative AI (GenAI) is at the center of the next wave of digital and data transformation. Businesses are increasingly looking to embrace the power of generative AI, a shift that often requires training models on proprietary data. This raises natural concerns about data sovereignty, security and cost.
In response, on-premises deployments using large open source language models like Llama 2 are gaining popularity. They offer predictable costs, control over data and reduced security and IP leak risks, ensuring regulatory compliance. Dell partners with Meta to provide businesses with better support for Llama 2.
At the same time, the transition from AI proof of concept to production can be complex and difficult for many organizations. That’s where Dell’s recently announced collaboration with Hugging Face comes into play.
Dell and Hugging Face work together on this process by providing a platform where businesses can easily select, deploy and fine-tune AI models for their specific use cases using Dell’s infrastructure.
Collaboration with cuddly faces
Dell is working with Hugging Face to simplify the creation, fine-tuning and implementation of open source generative AI models for businesses. This partnership aims to leverage Dell’s industry-leading infrastructure products and services with resources from the Hugging Face community.
A key part of this collaboration is a new Dell portal on the Hugging Face platform. This portal will allow Dell customers to easily choose from a library of open source AI models optimized for specific performance and use cases.
The portal offers custom containers and scripts to deploy these templates to Dell servers and data storage systems. The portal also provides access to datasets, libraries, and tutorials for training generative AI models, as well as templates and models for specific outcomes. Businesses can integrate them with their proprietary data and tailor it to their needs.
The fine-tuning process can be complex and time-consuming, using techniques such as Recovery Augmented Generation (RAG), LoRA and QLoRA. Dell aims to simplify this process with containerized tools.
Customers will initially have access to various Dell PowerEdge servers designed for AI models. Support for Dell workstations and Dell APEX service will be available soon.
This collaboration represents an important step in making open source AI more accessible and secure for enterprise applications, by combining the infrastructure strengths of Dell and the AI resources of Hugging Face.
Meta’s Llama in Dell’s Validated Design for GenAI
Dell also recently unveiled its Dell Validated Design for Generative AI, providing pre-tested hardware and software specifically for GenAI projects. Dell recently extended its validated design with a collaboration with Meta to facilitate on-premises deployment of Meta’s Llama 2 AI models using Dell infrastructure.
Dell Validated Design for Generative AI with Llama 2 delivers pre-tested and proven infrastructure, streamlining the deployment and management of on-premises GenAI projects. Llama 2, free for research and commercial use, is tested and verified on this Dell design, providing guidance for deployment and configuration. This allows for quick setup and predictable operation of Llama 2, especially for fine-tuning on Dell platforms.
The Dell PowerEdge XE9680 server, ideal for deploying Llama 2, is notable for being the first to ship with eight NVIDIA H100 GPUs and NVIDIA AI software. Llama 2 can also be used with various Dell infrastructures, such as the PowerEdge R760xa.
Dell also explored model customization in GenAI, demonstrating the application of customization techniques to Llama 2 models. This includes deploying the Llama 2 70B model on a PowerEdge XE9680 with Ubuntu 22.0 and NVIDIA GPUs, allowing for achieve efficient and accurate AI outputs.
Additionally, Dell is integrating Llama 2 models into its internal tools to help customers select the right solutions for their AI needs. This comprehensive approach provides organizations with reliable GenAI solutions, from desktops to data centers and clouds, ensuring efficient deployment and management of GenAI projects on-premises.
The analyst’s point of view
The relationship with Meta is important. Sy Choudhury, Meta’s vice president of AI partnerships, was at a recent AI-focused Dell analyst event in Austin. When asked what’s in that relationship for Meta, he said, “We need to make sure there’s an open path to generative AI.” »
This is similar to how Meta enables an open ecosystem to alleviate its infrastructure supply chain challenges through its strong participation in the Open Compute Foundation (which Dell also participates in).
The Meta relationship also makes sense for Dell. Almost every other LLM vendor competes with Dell in one way or another. Meta’s user-friendly approach will ensure an open ecosystem around large language models. Dell’s Hugging Face relationship follows the same path.
Dell’s competitors each take a different approach. With a relentless focus on its GreenLake as-a-service offerings, Hewlett Packard Enterprise offers an LLM-as-a-Service model with its GreenLake for LLM. Lenovo builds full-stack AI reference designs around its ThinkSystem offerings. Yet so far the company is going it alone without any announced GenAI model partnerships (aside from its relationship with NVIDIA, which almost every OEM maintains).
Generative AI is already transforming enterprise IT, but it is a complex problem. Having a technology partner like Dell that provides not only optimized and validated hardware solution stacks, but also the right set of core models and optimized tools, makes GenAI adoption easier for enterprise businesses. all sizes.
Most of today’s generative AI efforts are happening in the cloud. This only makes sense when considering the early stage of experimentation for many GenAI projects, coupled with the scarcity and cost of GPUs and accelerators needed for training and inference. As generative AI evolves, many of these workloads are naturally deployed on-premises. This is the now-normal application lifecycle of the modern cloud era. This is also where Dell is ready.
Dell’s collaborations with Meta and Hugging Face, as well as its recent organizational changes to support enterprise AI, tell us Dell is serious about helping customers adopt generative AI quickly and painlessly. Anything Dell does around GenAI will immediately benefit Dell’s customers, and anything that makes an IT worker’s life easier is a good thing.
Disclosure: Steve McDowell is an industry analyst and NAND Research, an industry analyst firm, which engages or has engaged in research, analysis and consulting services to numerous technology companies, which may include those mentioned in this article. Mr. McDowell has no stock ownership in any of the companies mentioned in this article.