Code Ahead of the Curve with FREE Technical Trainings

Retrieval-augmented generation is a technique for improving the accuracy of large language models by combining information retrieval from proprietary data sources with text generation.

This session explores how to implement the RAG architecture, a powerful framework for the RAG technique that integrates the strengths of open LLMs—such as Llama 3—and vector databases to improve contextual relevance and accuracy of information retrieval.

You’ll learn how to:

• Use open source models and tools to create a robust retrieval system that can efficiently process and generate natural language responses.
• Set up a vector database to efficiently store and retrieve embeddings.
• Integrate RAG systems into existing infrastructure, including the hardware and software requirements for deploying these advanced models.
• Apply best practices for fine-tuning and customizing RAG models and vector databases for specific industry use cases through demos of practical applications.

Sign up today.

Skill level: Intermediate

Featured software platforms

Session demos are done on the Intel® Tiber™ Developer Cloud¹, a managed cloud environment for development efficiency, cost savings, and faster time-to-market. Learn more & sign up

1 Formerly Intel® Developer Cloud
Intel, the Intel logo and Intel Tiber are trademarks of Intel Corporation or its subsidiaries.

Jani Janakiram
Principal Analyst, Advisor & Architect, Janakiram & Associates

The Intel Gaudi 2 AI accelerator has emerged as a game changer for AI compute, built from the ground up to accelerate generative AI and large language models. (Per the latest MLPerf Training v4.0 benchmark, it remains the only MLPerf-benchmarked alternative to the Nvidia H100 for training and inference.¹)

This session unpacks 3 ways to get started with Gaudi 2 to supercharge your LLM-based applications with performance, productivity, and efficiency, so you can:

• Run LLMs with Hugging Face Transformer models for training.
• Run inference on GenAI models using Hugging Face Optimum Habana.
• Migrate your models from a GPU environment to Gaudi 2 with just a few lines of code.

Sign up today.

Skill level: Intermediate

Get the software

• Gaudi Docker image on the Habana Vault contains all the software needed to run your models.

Get the code

• Gaudi Model References
• Hugging Face Optimum Habana
• GPU Migration Toolkit

1 Intel® Gaudi® 2 Enables a Lower Cost Alternative for AI Compute and GenAI
Intel, the Intel logo and Gaudi are trademarks of Intel Corporation or its subsidiaries.

Greg Serochi
Principal AI & Ecosystem-Enabling Manager, Intel® Gaudi® Accelerators

According to a myriad of trusted tech-focused websites, Intel® Arc™ GPUs offer great price-to-performance value for game development, media creation and, yes, generative AI.

This session focuses on how to implement high-performing GenAI applications on these daily-driver GPUs using Stable Diffusion (SD), Llama 3 quantization, and Intel-optimized extensions for PyTorch and Transformers. You’ll also see a demo on building a chatbot using the RAG technique.

Key learnings:

• How to deploy SD and large language models on Intel Arc GPUs.
• How to set up your own GenAI application with a few lines of code.
• How to quantize Llama 3 with INT4 via Intel AI-optimized software.
• How to build a chatbot with a RAG engine.

Sign up today.

Skill level: Novice

Get the software

• Intel® Extension for PyTorch
• Intel® Extension for Transformers

Intel, the Intel logo, and Arc are trademarks of Intel Corporation or its subsidiaries.

Zhang Jianyu
AI Software Solutions Engineer, Intel

Kai Lawrence Wang
AI Software Solutions Engineer, Intel

Federated learning (FL) is a machine learning approach for training models on decentralized edge devices without sharing raw data. Within this framework, organizations can collaborate on model development, models can gain experience from a vast range of data located at different sites, and data privacy can be preserved.

But FL implementation comes with challenges.

This session addresses one of them: Optimizing FL workload performance on CPUs and GPUs.

It leverages the implementation and results of a recent ASUS FL solution for the healthcare industry that significantly boosts efficiency and performance by using Intel® Xeon® processors with Intel® oneAPI and AI tools.

Key learnings:

• An overview of the ASUS solution—the approach, methodologies, and results.
• Using Intel AI software stacks and tools to analyze and enhance the performance of CNN models, e.g., VGG19 and EfficientNet, in real medical scenarios.
• Reviewing performance issues of AI framework and verifying the compatibility of the desired hardware.
• Collaboratively training a CNN model without sharing private data and images, using the open source FL framework Flower.

Sign up today.

Skill level: All

Featured software

• Intel® oneAPI Deep Neural Network Library standalone or as part of the Intel® oneAPI Base Toolkit
• Intel® VTune™ Profiler standalone or as part of the Intel oneAPI Base Toolkit
• Intel® Distribution for Python
• Intel® Extension for PyTorch

More resources

• ASUS – Unleashing the Power of Intel CPUs in Federated Learning: A Comprehensive Evaluation [pdf]

Intel, the Intel logo and VTune are trademarks of Intel Corporation or its subsidiaries.

Joel Lin
Software Technical Consulting Engineer

Joseph Lu
Assoc. Director Infrastructure Solution Group, ASUSTek

Dr. Lien Chungyueh
Assoc. Professor, Dept of Info Management, NTUNHS

The OpenVINO toolkit streamlines development, integration, and deployment of performant DL models in domains like computer vision, LLMs, and GenAI.

But given the ubiquity of heterogeneous compute environments, there’s a good chance your models and model-based apps must run on multiple hardware targets. Optimizing for each can take a bit of sleuthing.

This session addresses that issue, showing you how to configure and run analysis on your OpenVINO workloads to uncover bottlenecks on target hardware—CPU, GPU, NPU—using Intel® VTune™ Profiler and Intel® Advisor

Topics covered:

• The OpenVINO framework to boost AI application performance.
• Using Intel Advisor for performance modeling and CPU/GPU roofline generation.
• Using VTune Profiler for performance analysis across CPU (memory bottlenecks), GPU (Xe Vector Engine (XVE) utilization issues), and NPU (memory bandwidth).
• An overview of VTune’s native Instrumentation and Tracing Technology API (ITT API) utilities provided with the OpenVINO framework.

Sign up today.

Skill level: Intermediate

Featured software

• OpenVINO toolkit
• Intel VTune Profiler, standalone or as part of the Intel® oneAPI Base Toolkit
• Intel Advisor, standalone or as part of the Intel oneAPI Base Toolkit

Intel, the Intel logo, OpenVINO, and the OpenVINO logo are trademarks of Intel Corporation or its subsidiaries.
Intel, the Intel logo and VTune are trademarks of Intel Corporation or its subsidiaries.

Rupak Roy
Software Technical Consulting Engineer, Intel

Cory Levels
Software Technical Consulting Engineer, Intel