Google developers behind Swift for TensorFlow, which tunes the Apple-designed Swift programming language for machine learning applications, shared project guide data in an ongoing talk. Future plans for Swift for TensorFlow incorporate capacities, for example, improved automatic differentiation, C++ interoperability, and backing for distributed training.

CSS code on a screen
Photo by Sai Kiran Anagani / Unsplash

Swift for TensorFlow is a beginning period, Google-led project that incorporates Google's TensorFlow machine learning library with Swift, the advanced broadly useful language created by Apple. The utilization of Swift empowers all the more algorithms to be communicated in a new way, and simple differentiation of functions by means of generalized differentiation APIs, as indicated by the Swift for TensorFlow developers.

Open source Swift has been portrayed on the Swift for TensorFlow project site as simple to utilize and rich, with points of interest, for example, a powerful type system, which can enable developers to catch errors earlier and advances great API design. Expanding on TensorFlow, Swift for TensorFlow APIs give straightforward access to low-level TensorFlow administrators.

credit: Twitter

Swift for TensorFlow is centered around two set of users: propelled specialists limited by current machine learning systems, and machine learning students simply beginning. Extensions to the Swift language give interoperability among Swift and Python, a well known language in machine learning. Python can be imported inside a Swift Jupyter Notebook and TensorFlow itself is Python-accommodating. Developers can compose Swift to call into Python libraries, without any wrappers and no extra overhead.

You can download Swift for TensorFlow from GitHub. You can go to tensorflow.org/swift to find documentation, tutorials and guidelines for community participation in the project.