TensorFlow Development Service is typically a Python friend, although the real workreihe, implemented in C++ and CUDA, is performed. TF was released in 2015, and it has become ubiquitous: in research laboratories, in industry, in school. Approximately three times as many deep-learning experts use it as the second popular framework, and extensions such as TF-encrypted to do federated learning securely or Magenta to do AI-created art are available.