The English Version of this book is still under construction.
This is the corresponding code of the book PyTorch: Introduction and Practice. It can be used as independent PyTorch tutorials without the book. It contains several interesting PyTorch Projects. The Code is currently on PyTorch 1.0, with both CPU/GPU and Python2/3 compatibility.
The contents of the repo are shown in the figure:
It contains two parts:
Introduction (left of the mindmap). This part introduces the main modules of PyTorch and some tools commonly used in deep learning. Jupyter notebook is used as a teaching tool, you can modify and run the notebook interactively.
Chapter 2: Learning environment setup and a quickstart tutorial. You may spend about 1 to 2 hours to quickly complete the quickstart, and then choose to read the following related chapters in depth according to the needs.
Chapter 3: Introduction to the multidimensional array object (
Tensor) and dynamic graph engine (
autograd). Implement linear regression using Tensor and autograd respectively, and compare the differences between them. This chapter also makes a more in-depth analysis of the underlying design of tensor and autograd.
Chapter 4: Introduction to
torch.nn, and explanation of the common layers, loss function, optimizer and so on.
Chapter 5: introduction to the data loading, GPU acceleration, serialization and visualization tools.
Practice (Right of the mindmap). This part uses PyTorch along with other tools comprehensively to implement several cool and interesting projects.
Chapter 6: a connecting chapter to review PyTorch tools and apply it to a relatively easy task: image classification. In the process of implementation, guide the reader to review the knowledge of the first five chapters, and put forward the code specification to reasonably organizing the program and code so that the program is more readable and maintainable. This chapter also introduces how to debug in PyTorch.
Chapter 7: popular GAN for the readers, and guides you to implement an animate image generator from scratch.
Chapter 8: the knowledge of style transfer and guide you to implement the fast neural style, turning your photos into masterpieces.
Chapter 9: the knowledge of natural language processing and CharRNN. By collecting tens of thousands of Tang poems, you can train a small network that can write poems.
Chapter 10: Knowledge of image caption and takes the data of AI Challenger competition as an example to guide you to implement a small program that can carry out simple image description.
Licensed under MIT License.