Scribd
Upload
32 views2 pages

Hugging Face Case Study 112023

Uploaded by

Shyam Sunder Rao Chepur
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
32 views2 pages

Hugging Face Case Study 112023

Uploaded by

Shyam Sunder Rao Chepur
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 2

Case Study

Intel® Xeon® Processors


Habana® Gaudi®2 HPUs

Speed AI Development and Deployment Using


4th Gen Intel® Xeon® Processors, Habana® Gaudi®2
HPUs, and Hugging Face Open Source Libraries

AWS instances featuring Intel® AI acceleration technologies, with Optimum Intel


and Optimum Habana libraries, give companies powerful tools for generative
AI implementation.

Solution Summary Executive Summary


• Intel® Xeon® Processors Hugging Face is the leading open platform for AI builders. Its mission is to
democratize good machine learning through open science and open source,
• Intel® Advanced Matrix
including the Optimum Intel library, an extension of the Hugging Face
Extensions (Intel® AMX)
Transformers library. From a developer perspective, Hugging Face is the go-to
• Habana® Gaudi®2 HPUs place for free resources, including over a million open source models, data sets,
and applications that make generative AI and large language model development
• Intel® Extension for PyTorch easier. The combination of Hugging Face’s tools and AI-acceleration features
• OpenVINO™ toolkit built into 4th Gen Intel® Xeon® processors underlying Amazon EC2 instances
prove ideal for companies seeking a more turnkey approach for developing
• Amazon EC2 M7i, M7i-flex, performant and scalable AI solutions. Compared with 3rd Gen Intel Xeon
and C7i instances processors, the latest CPUs featuring Intel® Advanced Matrix Extensions
(Intel® AMX) can deliver 3x to 10x higher inference and training performance.1
Hugging Face also performed benchmark tests on the Habana® Gaudi®2 Habana
Processing Units (HPUs), finding them roughly twice the speed of Nvidia A100
80GB processors for training and inference.2

Challenge
While Hugging Face’s offerings simplify and streamline the process of developing
AI applications, companies often face another obstacle to deployment­—their
infrastructure. Running AI models in-house requires compute architectures
capable of accommodating heavy workloads. Organizations needed an easier
way to access the latest Intel hardware and software in the cloud to optimize their
generative AI and LLM implementations.

Hugging Face’s offerings simplify and streamline the process of developing AI


applications that companies often face.
Case Study | 4th Gen Intel® Xeon® Processors, Habana® Gaudi®2 HPUs, and Hugging Face Open Source Libraries Speed AI Development and Deployment

Solution Key Takeaways


In partnership with Intel, Hugging Face created the Optimum • Don’t try to create an AI model from scratch. Simplify
Intel library. The library makes the latest Intel Xeon processor the process using an applicable, pre-trained model from
hardware and software available to any Hugging Face user. Hugging Face.
When working with Amazon EC2 M7i, M7i-flex, and C7i
• Experiment with models and determine their
instances with 4th Gen Intel Xeon processors, users can
effectiveness using the free Hugging Face API.
benefit from software tools like the Intel® Neural Compressor
and built-in accelerators like Intel® Advanced Matrix • Take advantage of advanced technologies like Habana
Extensions (Intel® AMX). The OpenVINO™ toolkit also eases Gaudi2 in the Intel Developer Cloud.
generative AI deployments with high performance inference
optimization choices for PyTorch users. • Hugging Face’s Parameter Efficient Fine-Tuning
(PEFT) library can save users significant time when
Users planning to implement large language models (LLMs) fine-tuning language models.
with a billion or more parameters will appreciate technologies • Deploy models with APIs to build upon, like the Hugging
produced by Habana, an Intel company. The custom- Face Inference Endpoints service.
designed Habana Gaudi2 deep learning accelerator is
available through the Intel® Developer Cloud. Hugging Face
and Intel also offer the Optimum Habana library.
Where to Get More Information
“Amazon instances featuring 4th Gen Intel Xeon Explore Intel Xeon processors.
processors or Habana Gaudi HPU accelerators—
combined with the Hugging Face Optimum Read more about Habana Gaudi2 HPUs.

Intel and Optimum Habana open source Learn about Intel AMX.
libraries—provide incredibly efficient solutions for
Explore Hugging Face resources.
companies deploying AI models of all sizes.”
– Jeff Boudier, product director, Hugging Face Read about best practices for Amazon EC2 instances.

Results
Users adopting Amazon instances featuring 4th Gen
Intel Xeon processors with the Optimum library can gain a
significant performance boost. Testing found that CPUs
with Intel AMX can deliver 3x to 10x higher inference and
training performance than the previous generation of Intel
Xeon processors.1 Performance advantages like these help
users deploy AI solutions faster and more cost-effectively.
Benchmark tests also found Habana Gaudi2 processors
about twice as fast as Nvidia A100 80GB GPUs for both
training and inference.2

1
https://www.intel.com/content/www/us/en/products/docs/accelerator-engines/advanced-matrix-extensions/ai-solution-brief.html
2
https://huggingface.co/blog/habana-gaudi-2-benchmark

Performance varies by use, configuration and other factors. Learn more at www.Intel.com/PerformanceIndex.
Performance results are based on testing as of dates shown in configurations and may not reflect all publicly available updates. See backup for configuration details. No product or compo-
nent can be absolutely secure.
For workloads and configurations visit www.Intel.com/PerformanceIndex. Results may vary.
Intel does not control or audit third­party data. You should consult other sources to evaluate accuracy.
Your costs and results may vary.
Intel technologies may require enabled hardware, software or service activation.
© Intel Corporation. Intel, the Intel logo, and other Intel marks are trademarks of Intel Corporation or its subsidiaries. Other names and brands may be claimed as the property of others.

1023/RJMJ/JS/PDF Please Recycle 357430­001US

You might also like

pFad - Phonifier reborn

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy