Core generative models and techniques

Microsoft via Coursera

Go to class Write review

Overview

Coursera Flash Sale

40% Off Coursera Plus for 3 Months!

Grab it

Explore the diverse and powerful world of core generative AI. This course provides a comprehensive survey of the fundamental models that power modern AI, including Generative Adversarial Networks (GANs), autoregressive models, and diffusion models. You will build a strong foundation, understanding the unique architectures and training strategies for each, and compare essential frameworks like PyTorch and TensorFlow. The course then moves into hands-on implementation. You will learn to generate sequential data, such as time-series forecasts, using advanced autoregressive models in Azure AI Foundry. Next, you will master the art of high-fidelity image generation, using diffusion models to create and edit stunning visuals with techniques like inpainting and outpainting. Finally, you will learn to accelerate your development workflow by using Azure ML Designer, a visual, low-code environment for rapid prototyping. You will practice designing, building, evaluating, and preparing sophisticated model pipelines for real-world deployment. This course equips you not just with knowledge of different models, but with the practical skills to build and prototype them effectively on Azure.

Syllabus

Understanding core generative models

This foundational module introduces the diverse landscape of core generative models beyond LLMs. You will explore the distinct architectures and principles behind Generative Adversarial Networks (GANs), autoregressive models, and diffusion models. You will also dive into the practical aspects of model creation by comparing essential training frameworks like PyTorch and TensorFlow and learning the fundamental strategies for training these powerful models on Azure. Important Notice on the Azure Interface: The screencast videos and screenshots were last updated in late 2025. Please be aware that Microsoft may have updated the Azure interface since then. If the steps shown in the course materials look different from your current Azure environment, please follow the most up-to-date interface, as the underlying concepts and learning objectives remain the same.

Implementing autoregressive models

This module provides a deep dive into autoregressive models, the engines behind sequential data generation. You will focus on their application in tasks like time-series forecasting and text generation. Starting with the basic principles of next-token prediction, you will use Azure AI Foundry to implement models like TimeGEN-1. You will then advance to sophisticated techniques for controlling model output, ensuring your generated sequences are both coherent and high-quality. Important Notice on the Azure Interface: The screencast videos and screenshots were last updated in late 2025. Please be aware that Microsoft may have updated the Azure interface since then. If the steps shown in the course materials look different from your current Azure environment, please follow the most up-to-date interface, as the underlying concepts and learning objectives remain the same.

High-fidelity image generation with diffusion models

This module focuses on the cutting-edge technology of diffusion models for creating and editing stunning, high-fidelity images for any purpose. You will learn the fundamental "denoising" process that allows these models to generate photorealistic visuals—from creative compositions to professional graphics—using simple text prompts. You will then move beyond basic generation to master advanced techniques like inpainting, outpainting, and using negative prompts to gain precise control over your visual outputs. This will equip you to produce tailored, high-quality images for a wide array of business and creative applications. Important Notice on the Azure Interface: The screencast videos and screenshots were last updated in late 2025. Please be aware that Microsoft may have updated the Azure interface since then. If the steps shown in the course materials look different from your current Azure environment, please follow the most up-to-date interface, as the underlying concepts and learning objectives remain the same.

Foundations of ML pipelines with a visual designer

In this final module, we pivot from code-centric development to a powerful, high-level approach for accelerating model creation. You will master Azure ML Designer, a visual, drag-and-drop environment for rapid prototyping and pipeline development. You will learn to construct, train, evaluate, and prepare sophisticated models for deployment without writing extensive code. This module equips you with essential MLOps skills, enabling you to build and manage the entire machine learning lifecycle efficiently. Important Notice on the Azure Interface: The screencast videos and screenshots were last updated in late 2025. Please be aware that Microsoft may have updated the Azure interface since then. If the steps shown in the course materials look different from your current Azure environment, please follow the most up-to-date interface, as the underlying concepts and learning objectives remain the same.