PyTorch Official Tutorials
by PyTorch Contributors · PyTorch Foundation
The official PyTorch tutorial collection, organized from a Learn the Basics track through intermediate distributed training and advanced C++ extensions. Covers tensors, autograd, datasets, training loops, torch.compile, and export, so readers can build and optimize models against current APIs.
More resources on TensorFlow & PyTorch
Hands-On Machine Learning with Scikit-Learn, Keras, and TensorFlow (3rd Edition)
The standard reference for the TensorFlow/Keras side of deep learning, moving from scikit-learn regression and ensembles into Keras APIs, custom layers, tf.data, training loops, CNNs, transformers, and TFX deployment. Notebooks for every chapter are on GitHub.
PyTorch 2: Faster Machine Learning Through Dynamic Python Bytecode Transformation and Graph Compilation
The ASPLOS 2024 paper describing torch.compile: TorchDynamo rewrites Python bytecode to capture FX graphs, and TorchInductor lowers them to Triton and C++. Explains why eager-mode graph capture is hard and where compilation actually wins.
Deep Learning with PyTorch, Second Edition
A 2026 rewrite co-authored by PyTorch core developers, covering tensors, autograd, and the training loop before moving to transformers, diffusion models, and deployment. Ends with a full medical imaging project, so readers finish having shipped a non-toy model.
Keras 3 Developer Guides
Thirty-one official guides for Keras 3, which runs on TensorFlow, PyTorch, or JAX backends. Readers learn the functional API, layer subclassing, customizing fit(), backend-specific training loops, distributed training, quantization, and export, enough to write portable model code.
TensorFlow
TensorFlow.org is the official home of the TensorFlow machine learning framework. It provides comprehensive documentation, tutorials, API references, guides, and deployment resources for building, training, and deploying ML models across Python, mobile, and web platforms.