The most comprehensive collection of PyTorch learning resources in history

1. PyTorch learning tutorials and manuals

PyTorch English version official manual : For students who are good at English, I highly recommend the PyTorch official document, which will take you step by step from entry to proficiency. This document introduces in detail from basic knowledge to how to use PyTorch to build deep neural networks, as well as PyTorch syntax and some high-quality cases.
PyTorch Chinese official documentation : It doesn’t matter if you have difficulty reading the above English documentation. We have prepared a relatively official PyTorch Chinese documentation for you. The document introduces each function in great detail and can be used as a quick reference for PyTorch.
PyTorch code tutorial that is more partial to algorithm practice : It has a high star on github. It is recommended that you study the above two basic PyTorch tutorials before reading this document.
Open source book : This is an open source book that aims to help those who want to use PyTorch for deep learning development and research get started quickly. However, this document is not comprehensive and is still being updated.
Simple and easy-to-use PyTorch Chinese documentation : very suitable for novices to learn. This document starts from introducing what PyTorch is, to the installation of neural networks and PyTorch, to image classifiers, and data parallel processing. It introduces the knowledge system of PyTorch in great detail and is suitable for beginners to get started. The official website of this document: http://pytorchchina.com .

2. PyTorch video tutorial

PyTorch video tutorial at Bilibili : The first one I recommend is a PyTorch video tutorial with a very high click-through rate in Bilibili recently. Although the video content is only eight episodes, it is explained in simple terms and is very exciting. It’s just that there are no Chinese subtitles. It’s time for you guys to practice your English...
Foreign video tutorial : Another video tutorial by a foreign leader has a high click-through rate on YouTube. It is also a pure English video. Have you ever felt that foreign teaching videos can explain the problems very vividly and simply no matter how complex they are? ?
Mo Fan : I believe everyone should be familiar with Mr. Mo Fan. His series of videos on Python and deep learning have a high click-through rate on Bilibili and YouTube. This PyTorch video tutorial was just released last year. I recommend it to novice friends. .
101 Academy : The PyTorch series of video courses of the Artificial Intelligence 101 Academy are more detailed and cover a wide range of knowledge points. Interested friends can give it a try.
July Online : Finally, I recommend to you the PyTorch introductory and practical series of July Online, the leading artificial intelligence education platform in China. Although the course is a paid course, it includes PyTorch syntax, deep learning basics, word vector basics, NLP and CV project applications, practical combat, etc. The combination of theory and practical combat is indeed more detailed than other courses. It is recommended to everyone.

3. NLP&PyTorch actual combat

Pytorch text : Torchtext is a very easy-to-use library that can help us solve text preprocessing problems. This github repository contains two parts:
- torchText.data: Universal data loaders, abstractions and iterators for text (including vocabulary and word vectors)
- torchText.datasets: Pre-training loader for general NLP datasets. We only need to install torchtext through pip install torchtext, and then we can start to experience the convenience of Torchtext.
Pytorch-Seq2seq : Seq2seq is a rapidly evolving field, where new technologies and frameworks are frequently released. This library is a framework for the Seq2seq model implemented in PyTorch. The framework provides modular and extensible components for both training and prediction of the Seq2seq model. This github project is a base version with the goal of promoting these technologies and applications. development.
BERT NER : BERT is a pre-trained language model proposed by Google in 2018. Since its birth, it has broken a series of NLP tasks, so it has always had a very important influence in the field of nlp. This github library is the PyTorch version of BERT. It has many powerful pre-trained models built in, which is very convenient and easy to use.
An advertisement: If you need to activate the genuine PyCharm, you can contact me. It is 56 yuan per year. The genuine authorization is activated. The validity period can be checked on the official website. If you need it, add me on WeChat: poxiaozhiai6, remark: 914.
Fairseq : Fairseq is a sequence modeling toolkit that allows researchers and developers to train custom models for translation, summarization, language modeling, and other text generation tasks. It also provides reference implementations of various Seq2seq models. This github repository contains instructions for getting started, training new models, and extending Fairseq with new models and tasks. Friends who are interested in this model can click on the link above to learn.
Quick-nlp : Quick-nlp is a deep learning Nlp library heavily inspired by the fast.ai library. It follows the same API as Fastai and extends it to allow running NLP models quickly and easily.
OpenNMT-py : This is a PyTorch implementation of OpenNMT, an open source neural network machine translation system. It is designed to facilitate research, try out new ideas, and experiment with new ideas in many areas such as translation, summarization, image to text, morphology, and many more. Some companies have proven that this code can be used in real industrial projects, more details about this github can be found in the link above.

4. CV&PyTorch actual combat

pytorch vision : Torchvision is a library of handy tools for image manipulation independent of pytorch. It mainly includes: vision.datasets, vision.models, vision.transforms, vision.utils and several packages. The installation and use are very simple. Interested friends can refer to the above link.
OpenFacePytorch : This github library is an implementation of OpenFace in Pytorch, and the code requires that the input image be aligned and cropped in the same way as the original OpenFace.
TorchCV : TorchCV is a computer vision deep learning framework based on PyTorch, which supports the training and deployment of most visual tasks. This github library provides source code for most deep learning-based CV problems. Friends who are interested in the CV direction are still waiting. What?
Pytorch-cnn-finetune : This github library uses pytorch to fine-tune pre-trained convolutional neural networks. Supported architectures and models include: ResNet, DenseNet, Inception v3, VGG, SqueezeNet, AlexNet, etc.
Pt-styletransfer : This github project is neural style transfer in Pytorch. There are several specific points that need to be noted:
- StyleTransferNet as a class that can be imported by other scripts;
- VGG support (this was before pretrained VGG models were available in PyTorch)
- Ability to save intermediate styles and content targets for display
- Available as a function of the image inspection map matrix
- Automatic style, content and product image preservation
- Matplotlib plot of loss over time and hyperparameter logging to track favorable outcomes
Face-alignment : Face-alignment is a 2D and 3D face alignment library implemented in pytorch, using the world's most accurate face alignment network to detect facial landmarks from Python, capable of detecting points in 2D and 3D coordinates. This github library introduces in detail the basic process of face alignment using Face-alignment. Interested students are welcome to learn.

5. Recommended PyTorch papers

Google_evolution : This paper implements the resulting network that implements the large-scale evolution of image classifiers proposed by Esteban Real et al. Before the experiment, we need to install PyTorch, Scikit-learn and download the CIFAR10 dataset .
PyTorch-value-iteration-networks : This paper is based on the author's original Theano implementation and Abhishek Kumar's Tensoflow implementation, and includes the implementation of value iteration networks (VIN) in PyTorch. Vin won the best paper award at NIPS 2016.
Pytorch Highway : Highway Netowrks allows information to pass through each layer at high speed without hindrance. It is inspired by the gate mechanism in Long Short Term Memory (LSTM) recurrent networks. It allows information to pass through many layers without hindrance to train deep neural networks. The effect makes the deep neural network no longer only have the effect of the shallow neural network. This paper is an implementation of Highway network based on Pytorch.
Pyscatwave : Cupy/Pythorn scattering implementation. Scattering network is a kind of convolutional network whose filters are predefined as wavelets. It does not require learning and can be used for visual tasks such as image classification. Scattering transformation can significantly reduce the spatial resolution of the input (e.g. 224x224->14x14), and the pun power loss is significantly negative.
Pytorch_NEG_loss : This paper is a Pytorch implementation of Negative Sampling Loss. Negative Sampling is a method for solving the word2vec model. It abandons the Huffman tree and uses the Negative Sampling (negative sampling) method to solve it. This paper is a study of the loss function of Negative Sampling. Interested friends can Click on the paper link above to study.
Pytorch_TDNN : This paper is a Pytorch implementation of Time Delayed NN. The paper describes the principle and implementation process of TDNN in detail.

6. PyTorch book recommendations

Compared with the current situation where Tensorflow-type books are no longer available, there are not so many PyTorch-type books published. The author recommends four PyTorch books that I think are pretty good.

"Introduction to Deep Learning with PyTorch" , Electronic Industry Press, author: Liao Xingyu. This "Introduction to Deep Learning - PyTorch" is a relatively early one published among all PyTorch books. The author uses his own beginner's journey to deep learning to explain the syntax, principles and practical content of PyTorch in simple terms. It is suitable for Introductory learning for novices. But the disadvantage is that there are many loose and mechanical parts in the book, which need to be carefully distinguished by readers. Recommendation index: ★★★
"PyTorch Deep Learning" , People's Posts and Telecommunications Press, authors: Wang Hailing and Liu Jiangfeng. This book is an English translation. The original author is two Indian tycoons. In addition to the basic syntax and functions of PyTorch, the book also covers advanced neural network architectures such as ResNET, Inception, DenseNet, etc., as well as their application cases. . This book is suitable for readers such as data analysts and data scientists who have some theoretical foundation and practical experience. It is not recommended as an entry-level choice for novices. Recommendation index: ★★★
"Introduction and Practice of Deep Learning Framework PyTorch" , Electronic Industry Press, author: Chen Yun. This is a PyTorch book released in 2018. It contains two parts: theoretical introduction and practical projects. Compared with other books of the same type, this book has very detailed cases, including: classic projects in Kaggle competitions, GAN generated animation avatars, AI Filters, RNN poetry writing, image description tasks, etc. The theoretical + practical content setting is also more suitable for deep learning beginners and practitioners. Recommendation index: ★★★★
"PyTorch Machine Learning from Entry to Practical Practice" , Machinery Industry Press, author: Xiaobao Online, Sun Lin, etc. This book is also a Pytorch tutorial that combines theory with practice. Compared with the previous introductory + practical tutorial, the feature of this book is that the theoretical part about deep learning is very detailed, and the practical projects that follow are more comprehensive. Overall, this book is also a good introductory book to PyTorch suitable for novices. Recommendation index: ★★★