Overview

Coursera Plus Annual Sale: All Certificates & Courses 25% Off!

Grab it

NVIDIA: Large Language Models and Generative AI Deployment is the fourth course of the Exam Prep (NCA-GENL): NVIDIA-Certified Generative AI LLMs - Associate Specialization. This course offers a comprehensive understanding of Large Language Models (LLMs) and Generative AI deployment, combining theoretical insights with practical skills. Learners will explore key components of Generative AI, data requirements, and cleaning techniques for LLMs. The course covers model training, optimization, and evaluation methods, including Few-shot, Zero-shot, and Instruction Tuning. Additionally, the course dives into loss functions, alignment techniques, and evaluation metrics such as Perplexity. It also emphasizes the use of GPUs for training, fine-tuning methods like prompt tuning, and Parameter Efficient Fine Tuning (PEFT). Learners will gain expertise in LLM deployment strategies and monitoring with ONNX. This course is divided into three modules, each containing lessons and video lectures. Learners will engage with 4:30-5:00 hours of video content, covering both theoretical concepts and hands-on practices. Each module is equipped with quizzes to reinforce learning and assess understanding. Module 1: Fundamentals of Large Language Models Module 2: Training, Optimization, and Evaluation of LLMs Module 3: LLM Deployment Strategies and Monitoring By the end of this course, a learner will be able to: - Understand the foundational concepts of LLMs, including NLP and training data. - Explore model optimization techniques like loss functions, alignment, and PEFT. - Implement deployment strategies for LLMs and monitor performance using ONNX. This course is intended for professionals looking to deepen their expertise in deploying and optimizing LLMs for Generative AI applications.

Syllabus

Fundamentals of Large Language Models

Welcome to Week 1 of the NVIDIA: Large Language Models and Generative AI Deployment course. This week, we will begin by introducing you to Large Language Models (LLMs) and explore their significance in Natural Language Processing (NLP). We will also demonstrate how LLMs are applied to various NLP tasks using HuggingFace. Next, we will dive into the concept of Generative AI models and their components. We’ll cover the importance of training data for LLMs and best practices for data cleaning. By the end of this week, you will have a solid understanding of LLMs, their applications, and the essential processes involved in training them.

Training, Optimization, and Evaluation of LLMs

Welcome to Week 2 of the NVIDIA: Large Language Models and Generative AI Deployment course. This week, we will cover the essentials of training and optimizing Large Language Models (LLMs). We will begin by exploring the various learning methods, including Few-shot, Zero-shot, Instruction Tuning, and Reinforcement Learning with Human Feedback (RLHF). Next, we will delve into loss functions used in LLMs and techniques for aligning models effectively. We will also cover evaluation metrics such as Perplexity and discuss the critical role of humans in evaluating LLMs. Additionally, we will examine the role of GPUs in training models and explore LLM fine-tuning techniques like Prompt Tuning and Parameter Efficient Fine-Tuning (PEFT). By the end of the week, you will have a solid understanding of how to train, optimize, and evaluate LLMs for real-world applications.

LLM Deployment Strategies and Monitoring

Welcome to Week 3 of the NVIDIA: Large Language Models and Generative AI Deployment course. This week, we will cover essential strategies for deploying Large Language Models (LLMs) in real-world applications. We will start by exploring various deployment strategies and how to choose the right one for different scenarios. Next, we will introduce ONNX as a tool for unifying the deep learning landscape, and demonstrate how to convert deep learning models using ONNX. We will also focus on monitoring LLMs in production, covering best practices for ensuring their performance and reliability. Finally, we will dive into the NVIDIA ecosystem and how it supports LLM deployment, enhancing model efficiency and scalability. By the end of the week, you will have a clear understanding of LLM deployment and monitoring techniques.