19.5K ·GitHub
What is this?
The Chinese book 'Dive into Deep Learning' with its MXNet code examples rewritten in PyTorch — a free, open-source deep-learning tutorial pairing runnable Jupyter notebooks with online docs.
The Chinese book 'Dive into Deep Learning' with its MXNet code examples rewritten in PyTorch — a free, open-source deep-learning tutorial pairing runnable Jupyter notebooks with online docs.Is it a book I just read, or a library I add to my code?It's a textbook with its code: docs/ holds the chapters as markdown (served via docsify on GitHub Pages), code/ has a PyTorch notebook per chapter, and data/ holds the datasets.
It's a textbook with its code: docs/ holds the chapters as markdown (served via docsify on GitHub Pages), code/ has a PyTorch notebook per chapter, and data/ holds the datasets.How do I actually use it?Read it on the GitHub Pages site — easiest, since GitHub markdown won't render the book's LaTeX formulas. To run code, clone the repo and open the notebooks in code/. Offline reading works via docsify-cli or the 'd2dl' Docker image.
Read it on the GitHub Pages site — easiest, since GitHub markdown won't render the book's LaTeX formulas. To run code, clone the repo and open the notebooks in code/. Offline reading works via docsify-cli or the 'd2dl' Docker image.Do I need to know machine learning already to follow it?No — no deep learning or machine learning background is assumed. Basic linear algebra, differentiation, probability and Python are enough, and nothing needs installing to read the web version.
No — no deep learning or machine learning background is assumed. Basic linear algebra, differentiation, probability and Python are enough, and nothing needs installing to read the web version.Is the whole book finished, or are some parts still missing?Most chapters are complete, but several computer-vision sections are left unchecked in the README — SSD, semantic segmentation, FCN, style transfer and the CIFAR-10 / ImageNet Dogs Kaggle walkthroughs — so they look unfinished.
Most chapters are complete, but several computer-vision sections are left unchecked in the README — SSD, semantic segmentation, FCN, style transfer and the CIFAR-10 / ImageNet Dogs Kaggle walkthroughs — so they look unfinished.What topics does the book cover?Ten chapters: deep learning basics (linear regression, softmax, MLP, overfitting), model building and GPU computing, CNNs from LeNet to ResNet and DenseNet, RNNs with GRU and LSTM, optimizers up to Adam, plus computer vision (detection, segmentation, style transfer) and NLP (word2vec, seq2seq, attention, machine translation).
Ten chapters: deep learning basics (linear regression, softmax, MLP, overfitting), model building and GPU computing, CNNs from LeNet to ResNet and DenseNet, RNNs with GRU and LSTM, optimizers up to Adam, plus computer vision (detection, segmentation, style transfer) and NLP (word2vec, seq2seq, attention, machine translation).Does it include hands-on practice, or is it all theory?It's hands-on: every chapter ships runnable PyTorch notebooks, and the book walks you through real Kaggle competitions — house price prediction, CIFAR-10 image classification and ImageNet dog breed identification.
It's hands-on: every chapter ships runnable PyTorch notebooks, and the book walks you through real Kaggle competitions — house price prediction, CIFAR-10 image classification and ImageNet dog breed identification.What datasets does it use for exercises?Small classics you can train anywhere: Fashion-MNIST for image classification, a Pikachu dataset for object detection, and lyrics from Jay Chou albums for the language-model and RNN chapters.
Small classics you can train anywhere: Fashion-MNIST for image classification, a Pikachu dataset for object detection, and lyrics from Jay Chou albums for the language-model and RNN chapters.The original book also has an English edition — will this repo work for me?This repo follows the Chinese edition, so the text is in Chinese. The README notes the two editions differ and points readers who want a PyTorch version of the English edition to a separate repo.
This repo follows the Chinese edition, so the text is in Chinese. The README notes the two editions differ and points readers who want a PyTorch version of the English edition to a separate repo.What are similar projects?DeepLearning, pytorch-handbook, pytorch-tutorial, d2l-en, deep-learning-for-image-processing, nndl, deep-learning-with-python-notebooks, fastbook
DeepLearning, pytorch-handbook, pytorch-tutorial, d2l-en, deep-learning-for-image-processing, nndl, deep-learning-with-python-notebooks, fastbookWhich topics is it in?word2vec, kaggle, deep-learning-tutorial, textbook, cnn, attention-mechanism, neural-networks, natural-language-processing, jupyter notebook, book, computer-vision, learning-resources
word2vec, kaggle, deep-learning-tutorial, textbook, cnn, attention-mechanism, neural-networks, natural-language-processing, jupyter notebook, book, computer-vision, learning-resources