Dive Into DL PyTorch

Dive Into DL PyTorch (《动手学深度学习》PyTorch版) is a community project that rewrites the code in the Dive into Deep Learning textbook from MXNet into PyTorch, with the full Chinese text alongside it. The repository holds two folders: code, with a Jupyter notebook per section, and docs, with the book content in Markdown that is published as a website through docsify on GitHub Pages. The syllabus runs from environment setup, tensor operations and autograd through linear regression, softmax regression on Fashion-MNIST, multilayer perceptrons, weight decay and dropout, then into CNNs (LeNet, AlexNet, VGG, NiN, GoogLeNet, batch normalization, ResNet, DenseNet), RNNs (language models, GRU, LSTM, bidirectional and deep RNNs), optimization algorithms (SGD, momentum, AdaGrad, RMSProp, AdaDelta, Adam), computing performance including multi-GPU training, computer vision (image augmentation, fine-tuning, anchor boxes, SSD, R-CNN family, FCN, style transfer) and NLP (word2vec, fastText, GloVe, textCNN, seq2seq, beam search, attention, machine translation). Three Kaggle walkthroughs — house prices, CIFAR-10 and dog breed identification — close out the practical sections. You read it in the browser, or clone the repo and run the notebooks locally; a Docker image lets you serve the docs offline without installing Node.js. The project assumes only basic linear algebra, calculus, probability and Python. It is aimed at Chinese-speaking learners who want to learn deep learning in PyTorch rather than MXNet. The site states that chapters are still being added and that its content can differ slightly from the original book.