Google Open-Sources T2T, a Deep-Learning Trainer
Google continues to release new open-source artificial intelligence (A.I.) and machine learning (M.L.) tools. Its latest effort: Tensor2Tensor (T2T), a platform for training deep learning models in TensorFlow, an open-source library of machine-learning technologies. “T2T facilitates the creation of state-of-the art models for a wide variety of ML applications, such as translation, parsing, image captioning and more, enabling the exploration of various ideas much faster than previously possible,” Google mentioned in a corporate blog posting. This open-sourcing of T2T includes libraries of datasets and models. Google claims that T2T is noticeably faster and more resource-efficient when it comes to tasks such as machine translation. In terms of training time, for example, certain T2T models can accomplish quite a bit of training in a single day on a single GPU. Artificial intelligence and machine learning are such new and imposing concepts, it’s easy to feel a little intimidated when encountering them for the first time, especially when you expect that your job will need to incorporate some aspect of those disciplines in coming years. Fortunately, a number of companies have issued resources over the past year or so that allow anyone to begin wrapping their head around the core concepts. For example, Google offers a three-hour course in deep learning, a methodology for building effective machine-learning models. Available on the Google Cloud Platform Website, this course focuses on “a few basic network architectures, including dense, convolutional and recurrent networks, and training techniques such as dropout or batch normalization.” The course also includes an introduction to TensorFlow, which undergirds T2T. For its part, Facebook has a video series that breaks down modern A.I. platforms, including neural networks. It’s especially useful as an introduction to reinforcement learning, supervised learning, and predictive learning—key things to know for anyone seeking to break into the field. Whatever your level of knowledge, Google making these tools open-source is surely a big help when it comes to research or building products.