// the find
MorvanZhou/PyTorch-Tutorial
Build your neural network easy and fast, 莫烦Python中文教学
A grab-bag of short PyTorch scripts and matching notebooks covering regression, classification, CNNs, RNNs, autoencoders, GANs, and a DQN agent, from Chinese YouTuber Morvan Zhou's beginner tutorial series. Aimed at people who've never trained a network before and want a single-file example of each architecture rather than a course structure.
Each concept is one short, self-contained script (one file per architecture), so you can read 401_CNN.py top to bottom in a few minutes without chasing imports across a framework. Bilingual audience (Chinese/English) with matching video walkthroughs on the author's own site and YouTube, which is unusual reach for a tutorial repo. Covers the full spread a beginner actually asks about in one place — supervised, RNN, generative, and RL examples — instead of just classification.
Last commit is from March 2023 and the code predates PyTorch 0.4's Variable/Tensor merge in places, so some scripts use patterns (torch.autograd.Variable) that are deprecated and will confuse anyone following along on a current PyTorch install. No requirements.txt, environment file, or version pins anywhere, and no tests or CI. MNIST is committed as raw binary files in the repo (tutorial-contents/mnist/raw and processed), which bloats every clone instead of downloading via torchvision on demand. There's no narrative tying the scripts together — it's a flat list of examples, not a tutorial with progression or explanation beyond code comments.