thorough-pytorch

Why we included this project

Chinese-speaking learners who want to move from knowing Python to actually training models will find a coherent path here. The course is maintained by the Datawhale community and walks through tensor operations, autograd, building models, loss functions, and optimization, with each chapter pairing explanation with hands-on notebooks. Real projects like Fashion-MNIST classification give you something to run, and later chapters cover practical tricks such as adjusting learning rates, fine-tuning with torchvision or timm, half-precision training, and data augmentation. You also get into visualization with TensorBoard and wandb, the core PyTorch ecosystem libraries, and exporting models with ONNX. Because it is organized as a course instead of scattered snippets, it works both as a guided study plan and as a reference when you get stuck on your first real deep-learning project.