AI and ML for Coders in PyTorch Łaziska Górne

Eager to learn AI and machine learning but unsure where to start? Laurence Moroneys hands-on, code-first guide demystifies complex AI concepts without relying on advanced mathematics. Designed for programmers, it focuses on practical applications using PyTorch, helping you build real-world models without feeling overwhelmed. From computer vision and natural language processing (NLP) to generative AI with Hugging Face Transformers, this book equips you with the skills most in demand for AI development today. Youll also learn how to deploy your models across the web and cloud confidently. Gain the confidence to apply AI without needing advanced math or theory expertise Discover how to build AI models for computer vision, NLP, and sequence modeling with PyTorch Learn generative AI techniques with Hugging Face Diffusers and Transformers Spis treści: Foreword Preface Who Should Read This Book Why I Wrote This Book Navigating This Book Technology You Need to Understand Online Resources Conventions Used in This Book Using Code Examples OReilly Online Learning How to Contact Us Acknowledgments 1. Introduction to PyTorch What Is Machine Learning? Limitations of Traditional Programming From Programming to Learning What Is PyTorch? Using PyTorch Installing Porch in Python Using PyTorch in PyCharm Using PyTorch in Google Colab Getting Started with Machine Learning Seeing What the Network Learned Summary 2. Introduction to Computer Vision How Computer Vision Works The Fashion MNIST Database Neurons for Vision Designing the Neural Network The Complete Code Training the Neural Network Exploring the Model Output Overfitting Early Stopping Summary 3. Going Beyond the Basics: Detecting Features in Images Convolutions Pooling Implementing Convolutional Neural Networks Exploring the Convolutional Network Building a CNN to Distinguish Between Horses and Humans The Horses or Humans Dataset Handling the Data CNN Architecture for Horses or Humans Adding Validation to the Horses or Humans Dataset Testing Horses or Humans Images Image Augmentation Transfer Learning Multiclass Classification Dropout Regularization Summary 4. Using Data with PyTorch Getting Started with Datasets Exploring the FashionMNIST Class Generic Dataset Classes ImageFolder DatasetFolder FakeData Using Custom Splits The ETL Process for Managing Data in Machine Learning Optimizing the Load Phase Using the DataLoader Class Batching Shuffling Parallel Data Loading Custom Data Sampling Parallelizing ETL to Improve Training Performance Summary 5. Introduction to Natural Language Processing Encoding Language into Numbers Getting Started with Tokenization Using a custom tokenizer Using a pretrained tokenizer from Hugging Face Turning Sentences into Sequences Using out-of-vocabulary tokens Understanding padding Removing Stopwords and Cleaning Text Stripping Out HTML Tags Stripping Out Stopwords Stripping Out Punctuation Working with Real Data Sources Getting Text Datasets Getting Text from CSV Files Creating training and test subsets Getting Text from JSON Files Reading JSON files Summary 6. Making Sentiment Programmable by Using Embeddings Establishing Meaning from Words A Simple Example: Positives and Negatives Going a Little Deeper: Vectors Embeddings in PyTorch Building a Sarcasm Detector by Using Embeddings Reducing Overfitting in Language Models Adjusting the learning rate Exploring vocabulary size Exploring embedding dimensions Exploring the model architecture Using dropout Using regularization Other optimization considerations Putting It All Together Using the Model to Classify a Sentence Visualizing the Embeddings Using Pretrained Embeddings Summary 7. Recurrent Neural Networks for Natural Language Processing The Basis of Recurrence Extending Recurrence for Language Creating a Text Classifier with RNNs Stacking LSTMs Optimizing stacked LSTMs Using dropout Using Pretrained Embeddings with RNNs Summary 8. Using ML to Create Text Turning Sequences into Input Sequences Creating the Model Generating Text Predicting the Next Word Compounding Predictions to Generate Text Extending the Dataset Improving the Model Architecture Embedding Dimensions Initializing the LSTMs Embedding layers LSTM layers Final linear layer Variable Learning Rate Improving the Data Character-Based Encoding Summary 9. Understanding Sequence and Time Series Data Common Attributes of Time Series Trend Seasonality Autocorrelation Noise Techniques for Predicting Time Series Naive Prediction to Create a Baseline Measuring Prediction Accuracy Less Naive Predictions: Using a Moving Average for Prediction Improving the Moving-Average Analysis Summary 10. Creating ML Models to Predict Sequences Creating a Windowed Dataset Creating a Windowed Version of the Time Series Dataset Creating and Training a DNN to Fit the Sequence Data Evaluating the Results of the DNN Tuning the Learning Rate Summary 11. Using Convolutional and Recurrent Methods for Sequence Models Convolutions for Sequence Data Coding Convolutions Experimenting with the Conv1D Hyperparameters Using NASA Weather Data Reading GISS Data in Python Using RNNs for Sequence Modeling Exploring a Larger Dataset Using Other Recurrent Methods Using Dropout Using Bidirectional RNNs Summary 12. Concepts of Inference Tensors Image Data Text Data Tensors Out of a Model Summary 13. Hosting PyTorch Models for Serving Introducing TorchServe Setting Up TorchServe Preparing Your Environment Setting Up Your config.properties File Defining Your Model Creating the Handler File Creating the Model Archive Starting the Server Testing Inference Going Further Serving with Flask Creating an Environment for Flask Creating a Flask Server in Python Summary 14. Using Third-Party Models and Hubs The Hugging Face Hub Using Hugging Face Hub Getting a Hugging Face token Getting permission to use models Configuring Colab for a Hugging Face token Using the Hugging Face token in code Using a Model From Hugging Face Hub PyTorch Hub Using PyTorch Vision Models Natural Language Processing Other Models Summary 15. Transformers and transformers Understanding Transformers Encoder Architectures The self-attention layer The feedforward network layer Layer normalization Repeated encoder layers The Decoder Architecture Understanding token and positional encoding Understanding multihead masked attention Adding and normalizing The feedforward layer The linear and Softmax layers The Encoder-Decoder Architecture The transformers API Getting Started with transformers Core Concepts Pipelines Tokenizers The WordPiece tokenizer Byte-pair encoding SentencePiece Summary 16. Using LLMs with Custom Data Fine-Tuning an LLM Setup and Dependencies Loading and Examining the Data Initializing the Model and Tokenizer Preprocessing the Data Collating the Data Defining Metrics Configuring Training Initializing the Trainer Training and Evaluation Saving and Testing the Model Prompt-Tuning an LLM Preparing the Data Creating the Data Loaders Defining the Model Training the Model Managing data batches Handling the loss Optimizing for loss Evaluation During Training Reporting Training Metrics Saving the Prompt Embeddings Performing Inference with the Model The predict function Usage example Summary 17. Serving LLMs with Ollama Getting Started with Ollama Running Ollama as a Server Building an App that Uses an Ollama LLM The Scenario Building a Python Proof-of-Concept Creating a Web App for Ollama The app.js File The Index.html File Summary 18. Introduction to RAG What Is RAG? Getting Started with RAG Understanding Similarity Creating the Database Performing a Similarity Search Putting It All Together Using RAG Content with an LLM Extending to Hosted Models Summary 19. Using Generative Models with Hugging Face Diffusers What Are Diffusion Models? Using Hugging Face Diffusers Image-to-Image with Diffusers Inpainting with Diffusers Summary 20. Tuning Generative Image Models with LoRA and Diffusers Training a LoRA with Diffusers Getting Diffusers Getting Data for Fine-Tuning a LoRA Fine-Tuning a Model with Diffusers Publishing Your Model Generating an Image with the Custom LoRA Summary Index O autorze: Laurence Moroney pracuje w Google. Kieruje zespołem, który zajmuje się rozwiązaniami wykorzystującymi sztuczną inteligencję. Jest też trenerem: szkoli projektantów oprogramowania w zakresie technik budowy systemów uczenia maszynowego. Często udziela się na kanale TensorFlow w YouTube. Jest znanym na całym świecie prelegentem, a także autorem książek beletrystycznych — napisał kilka dobrze przyjętych powieści science fiction.