Model Optimization Foundational Principles

Overview

This course equips learners with essential techniques to enhance machine learning models. Starting with an introductory overview, the course covers key optimization strategies, including quantization techniques that reduce model size and improve efficiency. Students will explore pruning and sparsity methods to eliminate redundancy in models. The use of profiling tools and performance analysis is emphasized, allowing students to assess and refine their models effectively. Finally, the course culminates in practical applications, featuring hands-on experience with optimizing and deploying the GPT-2 model. Students will gain a solid foundation in optimizing state-of-the-art models for real-world applications.

Syllabus

Welcome to Model Optimization Foundational Principles

Get an introduction to key concepts and goals of model optimization in this course overview with a welcome video.

Introduction to Model Optimization

Explore model optimization, key metrics like latency, throughput, memory footprint, and perplexity, with hands-on demos, exercises, and quizzes for practical understanding.

Quantization Techniques

Explore quantization to shrink models, speed up inference, and lower costs by converting weights/activations to lower precision, using PTQ, QAT, and popular tools like Hugging Face Optimum.

Pruning and Sparsity

Learn how pruning removes unnecessary neural network weights for smaller, faster, energy-efficient models, covering unstructured and structured methods, tools, and quantization synergy.

Profiling Tools and Performance Analysis

Learn to analyze and optimize ML model performance using profiling tools like PyTorch Profiler and TensorBoard to find bottlenecks, tune data pipelines, and validate improvements.