In March this year, Google publicly demonstrated Swift for TensorFlow at the TensorFlow Developer Summit. Recently, the TensorFlow official website announced that Swift for TensorFlow is now open source.
Swift for TensorFlow provides a new programming model for TensorFlow that combines the TensorFlow computation graph with the flexibility and expressiveness of Eager Execution, while also focusing on improving usability at every layer of the entire software architecture.
The basis of the design is an algorithm called "Graph Program Extraction", which allows everyone to easily implement code using an Eager Execution-style programming model, while retaining the high-performance advantages of TensorFlow's computational graph.
The company emphasizes that Swift for TensorFlow is not just a wrapper around the TensorFlow API written in Swift, but includes compiler and language enhancements in Swift that benefit machine learning developers.
The design also supports advanced automatic differentiation in Swift and Python integration in Swift, allowing developers to use Python API Swift code directly.
From: sdtimes