PyTorch

PyTorch is a <a href="/facts/Machine_learning/e0w0XJTu">machine learning</a> <a href="/facts/Library_(computing)/xRzL3P7q">library</a> based on the <a href="/facts/Torch_(machine_learning)/2HyQQ2Mr">Torch</a> library, used for applications such as <a href="/facts/Computer_vision/Tl2Yyk66">computer vision</a> and <a href="/facts/Natural_language_processing/1hjMKsSN">natural language processing</a>, originally developed by <a href="/facts/Meta_AI/raPFo7MW">Meta AI</a> and now part of the <a href="/facts/Linux_Foundation/5wbe2Vp0">Linux Foundation</a> umbrella. It is one of the most popular <a href="/facts/Deep_learning/JLuwD3ea">deep learning</a> frameworks, alongside others such as <a href="/facts/TensorFlow/kDwfBeNd">TensorFlow</a>, offering <a href="/facts/Free_and_open-source_software/1cHtjLCB">free and open-source software</a> released under the <a href="/facts/Modified_BSD_license/XLtmLp5x">modified BSD license</a>. Although the <a href="/facts/Python_(programming_language)/YbuGqofa">Python</a> interface is more polished and the primary focus of development, PyTorch also has a <a href="/facts/C%252B%252B/Et0F2qn9">C++</a> interface.
A number of pieces of <a href="/facts/Deep_learning/JLuwD3ea">deep learning</a> software are built on top of PyTorch, including <a href="/facts/Tesla_Autopilot/1lS4a4Bf">Tesla Autopilot</a>, <a href="/facts/Uber/rHcJcp6S">Uber</a>'s Pyro, <a href="/facts/Hugging_Face/h8kkdUus">Hugging Face</a>'s Transformers, and Catalyst.
PyTorch provides two high-level features:

<ul><li>Tensor computing (like <a href="/facts/NumPy/T6FnhWWD">NumPy</a>) with strong acceleration via <a href="/facts/Graphics_processing_unit/PTK1RQVp">graphics processing units</a> (GPU)</li>
<li><a href="/facts/Deep_neural_networks/JLuwD3ea">Deep neural networks</a> built on a tape-based <a href="/facts/Automatic_differentiation/dQcdhBAM">automatic differentiation</a> system</li></ul>

PyTorch open-in-new

PyTorch