Organization

Organization#

This guide offers multiple learning paths for students, educators, and AI enthusiasts looking to explore AI on AMD platforms. Choose the path that best matches your goals and experience level.

Learning Paths#

We have structured this guide into three key learning paths:

  • Inference Path: Get started by running pre-trained models on AMD hardware.

  • Training Path: Train your own model or fine-tune an existing one with your custom dataset.

  • Optimization Path: Learn essential and advanced techniques to maximize model performance on AMD platforms.

Each path builds on the previous one, but you can jump in at any stage based on your needs.

Running Inference#

The inference path is designed for beginners who want to get started running AI models on AMD technology.

After you take this path, you will be able to:

  • Install dependencies to use PyTorch for AMD hardware.

  • Run a pre-trained model from Hugging Face or PyTorch Hub on AMD technology.

Start by navigating to the getting-started section.

Model Training#

The training path is designed for learners with some AI experience who want to train their own models or fine-tune existing ones.

After you take this path, you will be able to:

  • Generate or get data to train a model.

  • Define your data split to train, validate and test your model.

  • Define your own model with PyTorch.

  • Train a model on AMD technology.

  • Fine-tune a pre-trained model.

  • Explore AI explainability to understand why a model makes a decision.

Start by navigating to the training section.

Model Optimization#

This path is designed for users that want to maximize the performance for a given model on AMD hardware.

After you take this path, you will learn how to:

  • Quantize models for improved efficiency.

  • Develop and optimize HIP kernels.

  • Use Flash Attention for faster computations.

  • Profile applications to identify bottlenecks.

  • Fine-tune models for performance gains.

  • Apply AI explainability techniques to interpret model decisions.

Start by navigating to the optimization section.

Tutorials#

If you prefer a structured, step-by-step approach, the tutorials path guides you through the entire AI workflow from setup to deployment.

You will follow a linear progression:

  1. Set up your environment for AMD hardware.

  2. Run a pre-trained model to verify the setup.

  3. Define a model and prepare your dataset.

  4. Train the model from scratch or fine-tune an existing one.

  5. Optimize the model for efficient deployment.

  6. Deploy the model on AMD technology.

See Tutorials for more information.

Copyright (C) 2025 Advanced Micro Devices, Inc. All rights reserved.

SPDX-License-Identifier: MIT

Contents