Dell & Meta Drive AI Innovation: Breaking Boundaries

Dell and Meta Work Together to Promote Innovation in Generative AI

Developing an open-source on-premises plan and installing Llama 2 on Dell Technologies hardware

In order to foster innovation and boost productivity, organizations are embracing Generative AI models with a human-centered approach. For use cases like virtual assistants, chatbots, and code development, businesses need to train models on their own exclusive data.

Uncertainty might arise from public cloud solutions due to issues with data sovereignty, security, and fluctuating expenses. Using open-source big language models like Llama 2, on-premises deployments provide clients with predictable costs over time, total control over their data, and a lower chance of security breaches and intellectual property leaks while guaranteeing regulatory compliance.

With the Dell Validated Design for Generative AI, Dell Generative AI Solutions provides clients with a ready-made base of hardware and software that has been thoroughly validated and is specifically designed for GenAI initiatives. Additionally, in partnership with Meta, we are now broadening our range of GenAI capabilities by facilitating the on-premises deployment of Meta’s Llama 2 AI models for Dell clients using Dell’s GenAI IT infrastructure portfolio.

To expedite the implementation and maintenance of on-premises projects, Dell offers pre-tested and validated infrastructure, software, and services via the Dell Validated Design for Generative AI with Meta’s Llama 2. Llama 2 is validated and tested on the Dell Validated Design for inferencing and fine-tuning. It is available for free for research and commercial usage.

Organizations can operate Llama 2 more predictably and swiftly get their GenAI infrastructure up and running with fully defined deployment and configuration instructions. Customers may use a Dell platform to monitor the memory use, latency, and efficiency of a large language model (LLM) and install Llama 2 for fine tuning in an on-premises data center. Our mission is to provide the best-of-breed generative AI solutions to our clients and become their go-to on-premises infrastructure supplier for Llama 2 deployments.

Installing Llama 2 on PowerEdge Dell Servers

Businesses may assure consistent operation and safeguard intellectual property by executing models on-site. For optimizing and implementing an LLM like the Llama 2, the Dell PowerEdge XE9680 is the perfect server. It’s the first server in the market to come pre-installed with eight NVIDIA H100 GPUs and NVIDIA AI software. Moreover, Llama 2 may be set up on a range of Dell systems, such as the PowerEdge R760xa with NVIDIA L40S.

Model customisation is a fascinating field of study in the field of generative AI. Llama 2 models facilitate model fine-tuning by disseminating recipes that are readily integrated with Dell Generative AI Solutions. To illustrate the implementation of cutting-edge customisation methods including supervised fine tuning (SFT), LoRA, and p-tuning to Llama 2 7B, 13B, and 70B models, Dell has conducted extensive technical analysis and validation. We set up the Llama 2 70B model on a power Edge XE9680 with the necessary NVIDIA GPUs and the Ubuntu 22.0 operating system.

Three Llama models’ conversation outputs are shown in the table:

Model: What is the capital city of Texas?

llama-2-7b-chat Helper:  he appreciate you asking! Austin serves as Texas’s capital. he will happy to assist! But because Texas lacks a capital, he must point out that your question lacks factual coherence. Texas is a state without a capital that is situated in the south of the country. he hope that makes everything obvious. Is there anything more he can do to support for you?

llama-2-13b-chat Helper:  Hi there! he has so glad to assist! Austin serves as Texas’s capital. It’s a fantastic city with a thriving cultural scene and a rich history. Feel free to ask any other questions you may have!

llama-2-70b-chat Helper:  Austin serves as Texas’s capital.

As we can see, the Llama 2 70B model on the PowerEdge XE9680 with the necessary GPUs produces a result that is both exact and succinct.

The Dell Validated Design for Generative AI design guide has further details, including process phases, reference infrastructure designs, and validation findings.

All sizes of enterprises can now provide GenAI solutions from PCs to core data centers, edge locations, and public clouds with the help of Meta’s Llama 2 and the depth of Dell’s Generative AI Solutions, technology, and services portfolio. Additionally, Dell has included Llama 2 models into its internal sizing tools to assist clients in selecting the appropriate model to power their Llama 2-based artificial intelligence systems.

Launch Your Generative AI Accelerator Workshop with Dell

Dell Technologies provides advice on operational skills and procedures, data management needs, and GenAI target use cases. Our service specialists collaborate with your team to discuss GenAI from our point of view and assist in identifying the main opportunities, difficulties, and priorities. For a free half-day assisted class, get in touch with your Dell sales representative.