pytorch-handbook

Why we included this project

This is a Chinese-language open book that takes PyTorch learners from the first tensor to real projects, organized as a real curriculum rather than a loose pile of snippets. Early chapters cover environment setup, autograd, and the core neural network building blocks, then move into worked examples like MNIST digit recognition and an RNN that predicts cosine from sine. Later chapters go into fine-tuning, visualization with visdom and TensorBoardX, multi-GPU parallel training, and even a Raspberry Pi build, so the book stays useful well past the beginner stage. Every notebook is tested to run, which means learners can trust the code instead of debugging stale examples, and the version notes track PyTorch releases to keep the material current. Teams onboarding new members to deep learning, or individuals who want a structured Chinese-language reference, will find it a practical companion to the official documentation.