← Software

PyTorch

Open-source machine-learning framework with dynamic computation graphs.

PyTorch is the dominant deep-learning framework for research and increasingly for production. Models are defined as Python code executing against a tensor library backed by C++ and CUDA, with automatic differentiation for gradient-based training. Its define-by-run model means a network's structure can change at runtime, which made it the framework of choice for natural language processing, vision, and generative models throughout the late 2010s and 2020s.

PyTorch runs natively on Linux with CUDA, ROCm, or CPU backends, and is the substrate for Hugging Face Transformers, Stable Diffusion, Whisper, LLaMA, and most modern open-weight models.

pip install torch
conda install pytorch -c pytorch