PyTorch is an open-source machine learning and deep learning framework providing tensors with strong GPU acceleration and dynamic neural networks built in Python. At its core is an automatic differentiation engine (autograd) and an eager execution model that makes debugging and prototyping models straightforward.
The framework was originally created by Meta AI (formerly Facebook AI Research, FAIR) and, since September 2022, has been developed under the PyTorch Foundation, part of the Linux Foundation. It is written in Python, C++ and CUDA, and its source code is released under the BSD-3-Clause license.
PyTorch runs on Linux, Windows and macOS, and supports acceleration on NVIDIA (CUDA) and AMD (ROCm) GPUs as well as Apple Silicon and Intel GPUs. Version 2.0 (2023) introduced the torch.compile compiler built on TorchDynamo. In robotics, PyTorch is widely used as the backend for training and deploying perception, control and robot-learning models.
An API Library is a software package that exposes programmatic interfaces for communicating with a device, service, or system. In robotics it typically forms a lightweight integration layer built on top of the manufacturer's official API or an open-source project, abstracting low-level protocol details and providing language-native bindings (Python, C++, Java, etc.).
The robot learning role describes software for training a robot's control policies and manipulation/locomotion skills using machine learning methods. It covers: reinforcement learning in simulation with sim-to-real transfer, imitation learning and learning from demonstration, training Vision-Language-Action (VLA) models, and fine-tuning robotics foundation models. It typically uses massively parallelized simulation environments (Isaac Lab, MuJoCo) to generate training data, then deploys the trained policies on a physical robot.
Widely used in AI research and production, including at Meta, Tesla (Autopilot) and OpenAI, and as the backend of the Hugging Face Transformers library.
Over 102,000 stars and ~29,000 forks on GitHub; one of the most popular deep learning frameworks with a broad ecosystem (torchvision, torchaudio and more).
Ubuntu 24.04 LTS 'Noble Numbat' — supported until April 2029. The host for ROS 2 Jazzy.
x86-64 or ARM64 CPU. Officially tested on Ubuntu, Windows and macOS. The maintainers do not publish strict minimum RAM/storage requirements.
License family: Permissive
Latest stable release in the 2.x line (as of August 2026).
Introduced torch.compile built on TorchDynamo, significantly speeding up execution while remaining backward compatible.
First public release of PyTorch by Facebook AI Research (FAIR).
