PyTorch is an open-source deep learning framework and ecosystem maintained by the PyTorch Foundation. It provides tensor operations with strong GPU acceleration, a dynamic computation graph and automatic differentiation (autograd) that make prototyping and research fast and intuitive, plus tools for optimizing and serializing models (TorchScript) for production. PyTorch also offers a C++ frontend (libtorch), extensive community examples and tutorials, and a growing ecosystem of libraries (e.g., TorchVision, PyTorch Lightning, PyTorch Geometric) that extend functionality for vision, distributed training, graph learning, and model serving. Its combination of researcher-friendly APIs and production-oriented tooling helps teams iterate quickly and scale models to multi-GPU and multi-node environments.