Swift for TensorFlow project shuts down

Swift for TensorFlow project shuts down

Google pulls the plug on effort to integrate the Swift language and TensorFlow machine learning library, while citing multiple achievements

Credit: Dreamstime

Swift for TensorFlow, a Google-led project to integrate the TensorFlow machine learning library and Apple’s Swift language, is no longer in active development. Nevertheless, parts of the effort live on, including language-differentiated programming for Swift.

The GitHub repo for the project notes it is now in archive mode and will not receive further updates. The project, the repo notes, was positioned as a new way to develop machine learning models.

“Swift for TensorFlow was an experiment in the next-generation platform for machine learning, incorporating the latest research across machine learning, compilers, differentiable programming, systems design, and beyond."

Developers involved in the project and Google itself had not responded to InfoWorld inquiries about the fate of the project by early evening on February 16.

Nonetheless, proponents of Swift for TensorFlow cited in the repo a number of accomplishments which include the addition of language-integrated differentiated programming into Swift, with this work continuing in the Swift compiler.

Furthermore, the development of a mutable-value-semantics-oriented deep learning API and enabling of novel research combining deep learning with probabilistic graphical models for 3D motion tracking and beyond, via the SwiftFusion project.

This is alongside spinning off of multiple open source side efforts still in active development such as PythonKit, for Python interoperability with Swift; swift-jupyter, for using Swift within Jupyter notebooks, and swift-benchmark, for benchmarking Swift code.

Meanwhile, several Swift evolution proposals were also referenced, including:

  • SE-0195: User-defined “dynamic member lookup” types;
  • SE-0216: Introduce user-defined dynamically “callable” types;
  • SE-0233: Make numeric refine a new AdditiveArithmetic protocol;
  • SE-0253: Callable values of user-defined nominal types.

Developed by Google and released to open source in 2015, TensorFlow perhaps has lost some of its luster lately. Facebook’s PyTorch has been luring away TensorFlow users including Disney and Blue River Technology, which chose PyTorch over TensorFlow. Ease of use has been cited as a benefit of PyTorch.

Tags Google

Show Comments