This workshop will concentrate on the latest developments in Deep Learning. Participants will explore the optimisations made to Deep Learning frameworks like Tensorflow and PyTorch, and how these optimisations are integrated into their stock versions. Additionally, attendees will learn how to use Intel extensions to enable extra features like mixed-precision and running on Intel dGPU, using Stable Diffusion as an example of how these optimisations can be beneficial. The workshop will also emphasise inference optimisation tools such as Intel Neural Compressor and OpenVINO, which leverage graph optimisations and/or quantisation to accelerate Deep Learning inference.

Introduction to AI and Intel Software for AI:

• A short history of AI
• Current status of Machine Learning and Deep Learning
• Latest development in Deep Learning
• Overview of Hardware and Software by Intel 

Optimise the foundations of Deep Learning:

• Optimisations of the Deep Learning Frameworks
• Use mixed precision and Intel dGPU using the Intel Extensions

Optimise Stable Diffusion with Intel

An overview of Uncertainty Estimation techniques in AI 

Introduction to Neural Network Optimisation Techniques

• Compression and Optimisation Techniques Using the Intel Neural Compressor
• Graph optimisations using OpenVINO 

AI-driven multiphysics HPC applications on Intel architecture

Prerequisites:

• Basic knowledge of Python
• AI Training Series: Orientation Session (or comparable previous knowledge)