Code Ahead of the Curve with FREE Technical Trainings

MLPerf is a benchmarking suite that measures the real-world performance of machine-learning systems against a variety of ML tasks—image classification, object detection, machine translation, and others, doing so in an architecture-neutral, representative, and reproducible manner.

This session focuses on Intel’s results from its 2023 MLPerf v.3.0 submissions specific to thedata center category. The workload tasks were run on 4th Gen Intel® Xeon® Scalable processors. (Fun fact: Intel remains the only data center CPU vendor to have MLPerf Inference results on a broad set of models.)

Topics covered include:

• The remarkable gains of 4th Gen versus 3rd Gen Intel Xeon processors to run any AI workload, due in large part to specialized AI hardware accelerators like Intel® AMX
• How additional joint submissions with other customers (all on 4th Gen Xeon) were competitive to NVIDIA GPUs
• Key learnings to boost model performance on 4th Gen Xeon CPUs such as platform configuration, memory balancing, and recommended BIOS settings (these can be checked using Intel® System Health Inspector
• Methodologies and tools Intel used to optimize model performance for this submission
• Tips for how you can use MLPerf to benchmark your own model performance

Skill level: All

Featured software

Download the following standalone:

• Intel® Extension for PyTorch
• Intel® Neural Compressor
• Intel® oneAPI Deep Neural Network Library (oneDNN)

See the results and code

• Explore the GitHub repository for Intel’s MLPerf v3.0 Inference results.

Yuning Qiu
AI Software Solutions Engineer

Two problems continuously faced by machine learning engineers and data scientists are (1) efficiently managing AI pipelines from development to deployment and (2) running those pipelines in such a way as to reduce costs and resource use.

Microsoft and Intel have collaborated to create a solution that addresses both by incorporating Intel AI optimizations into the Azure Machine Learning platform. Sign up for this session to learn what it is and how to take advantage of it.

Key takeaways:

• An overview of the solution, which integrates Intel optimizations for Python frameworks such as PyTorch, TensorFlow, and Scikit-learn on the Azure ML platform
• How to enable in your Azure workloads using the above-mentioned framework optimizations to reduce cloud costs and development time, increase resource utilization, and improve ML pipeline speed

Includes a demo.

Sign up today.

Skill level: Intermediate

Featured software

Get the following Intel extensions standalone from GitHub or as part of the Intel® oneAPI AI Analytics Toolkit

• TensorFlow
• PyTorch
• Scikit-learn

Download code samples

• Intel TensorFlow – Getting Started
• Intel PyTorch – Optimizing Workloads
• Intel Scikit-learn – Getting Started
• See all code samples

Rachel Oberman
Intel AI Technical Consulting Engineer

Savita Mittal
Principal Software Engineer, Microsoft

This session addresses the challenge of speeding up the linear algebra oneMKL kernel on GPUs through OpenMP. (This is important because it allows you to maintain the same code base for CPU and GPU OpenMP acceleration.)

You’ll learn the technique via a demonstration of how to solve a 1000x1000 complex number matrix by calling the GPU version of the oneMKL subroutine on Intel® Data Center GPU Max Series processors.

Key takeaways:

• An overview of OpenMP, including how it’s a native solution for Fortran users to take advantage of Intel GPU hardware and its offload capabilities
• How to dispatch a oneMKL math kernel, such as solving a complex number matrix, to the latest Intel GPU
• Compiling the code example using the latest Intel® Fortran Compiler
• Using OpenMP data construct to manage data movement between the host and target devices
• How to monitor and optimize host-target data movement by the OpenMP runtime library

Sign up today.

Skill level: Intermediate

Featured software

Download the following standalone or as part of the Intel® oneAPI Base Toolkit

• Intel oneAPI Math Kernel Library
• Intel Fortran Compiler – Windows | Linux | macOS

Download code samples

• oneMKL – Solving Linear Systems
• Fortran-OpenMP Offload to GPU
• Intel Fortran Compiler Co-array Tutorial
• See all code samples

Shiquan Su
Software Technical Consulting Engineer

Barbara Perz
Technical Consulting Engineer

Regardless of whether medical science is your thing, this session is a good one for anyone interested in materially improving performance on Intel CPUs of computationally challenging, algorithm-heavy applications.

Sign up to hear how two researchers accomplished it for a critical epistasis disease-detection application that uses datasets spanning millions of genetic markers in multiple high-order combinations; this effectively expands datasets into the hundreds of trillions of samples that must be quickly and accurately evaluated.

Find out how they leveraged oneAPI tools to transform their legacy code into multi-architecture state-of-the-art code, plus the results they achieved on Intel CPUs, including Intel® Xeon® CPU Max Series with HBM.

This session focuses on Intel’s results from its 2023 MLPerf v.3.0 submissions specific to thedata center category. The workload tasks were run on 4th Gen Intel® Xeon® Scalable processors. (Fun fact: Intel remains the only data center CPU vendor to have MLPerf Inference results on a broad set of models.)

Sign up today.

Skill level: Intermediate

Featured software

Get the following tools standalone or as part of the Intel® oneAPI Base Toolkit

• Intel® oneAPI DPC++/C++ Compiler
• Intel® Advisor
• Intel® DPC++ Compatibility Tool

Download code samples

• Matrix Multiply with Intel Advisor
• Compatibility Tool : Migrate CUDA to DPC++/SYCL with Vector-Add
• See all code samples

Aleksandar Ilic
Assistant Professor, Universidade de Lisboa and Senior Researcher SIPS GROUP, INESC-ID

Ricardo Nobre
Researcher at INESC-ID | High-Performance Computing Architectures and Systems (HPCAS)