TACO is an open-source tensor algebra compiler that generates efficient code for dense and sparse tensor computations across a variety of storage formats. The project’s website serves as a hub for documentation, source code, and resources for using the compiler.