Deep Learning with PyTorch Step-by-Step

Revised for PyTorch 2.x!  Volume I: Fundamentals & Volume II: Computer Vision & Volume III: Sequences & NLP

Why this book?

Are you looking for a book where you can learn about deep learning and PyTorch without having to spend hours deciphering cryptic text and code? A technical book that’s also easy and enjoyable to read? This is it!

How is this book different?

• First, this book presents an easy-to-follow, structured, incremental, and from-first-principles approach to learning PyTorch.
• Second, this is a rather informal book: It is written as if you, the reader, were having a conversation with Daniel, the author.
• His job is to make you understand the topic well, so he avoids fancy mathematical notation as much as possible and spells everything out in plain English.

What will I learn?

In this first volume of the series, you’ll be introduced to the fundamentals of PyTorch: autograd, model classes, datasets, data loaders, and more. You will develop, step-by-step, not only the models themselves but also your understanding of them.

In this second volume of the series, you’ll be introduced to deeper models and activation functions, convolutional neural networks, initialization schemes, learning rate schedulers, transfer learning, and more.

In this third volume of the series, you’ll be introduced to all things sequence-related: recurrent neural networks and their variations, sequence-to-sequence models, attention, self-attention, and Transformers.

What’s Inside

• Gradient descent and PyTorch’s autograd
• Training loop, data loaders, mini-batches, and optimizers
• Binary classifiers, cross-entropy loss, and imbalanced datasets
• Decision boundaries, evaluation metrics, and data separability
• Deep models, activation functions, and feature spaces
• Torchvision, datasets, models, and transforms
• Convolutional neural networks, dropout, and learning rate schedulers
• Transfer learning and fine-tuning popular models (ResNet, Inception, etc.)
• Recurrent neural networks (RNN, GRU, and LSTM) and 1D convolutions
• Seq2Seq models, attention, masks, and positional encoding
• Transformers, layer normalization, and the Vision Transformer (ViT)
• BERT, GPT-2, word embeddings, and the HuggingFace library