Facilitating the Spread of Knowledge and Innovation in Professional Software Development

Write for InfoQ


Choose your language

InfoQ Homepage News Propel Shifts Plans to Leverage TensorFlow.js

Propel Shifts Plans to Leverage TensorFlow.js

This item in japanese


The Propel JavaScript scientific computing and machine learning library has announced a change in the project's direction. Just a few weeks after Propel's initial launch in March 2018, TensorFlow.js announced its release. Propel's initial efforts extended deeplearn.js and the C implementation of TensorFlow. Tensorflow.js is an evolution of deeplearn.js, a JavaScript library released by Google.

Given the similarity between TensorFlow.js and Propel's lower-level approach, the Propel project team quickly realized that it would be better to converge on a shared platform:

TensorFlow.js was recently released. It is well engineered, provides an autograd-style interface to backprop, and has committed to supporting Node. This satisfies our requirements. It is counterproductive to pursue a parallel effort. Thus we are abandoning our backprop implementation, TF C binding, and the TF/DL bridge, which made up the foundation of the Propel library. We intend to rebase our work on top of TFJS.

As such, the Propel project is currently rebooting. The website and examples for using Propel are no longer available, and public-facing activity within the project has been minimal over the past few weeks as the Propel team works to define their new direction:

Our high-level goal continues to be a productive workflow for scientific computing in JavaScript. Building on top of TFJS allows us to focus on higher-level functionality.     

Similar to the original early work on Propel, TensorFlow.js also leverages WebGL for GPU-supported numerical operations. According to the TensorFlow.js team, support for Node.js is now available:

Yes! We recently released Node.js bindings for TensorFlow. This allows the same JavaScript code to work on both the browser and Node.js, while binding to the underlying TensorFlow C implementation in node. You can follow its development on GitHub or try out our NPM package.

As part of the release of TensorFlow.js, the deeplearn.js library has become TensorFlow.js Core. Beyond the core library, TensorFlow.js adds a Layers API for building machine learning models and tools for automatically porting TensorFlow SavedModels and Keras HDF5 models.

Propel and TensorFlow.js are both open source projects under the Apache 2.0 license. Contributions are encouraged via GitHub repositories for TensorFlow.js and Propel.

Rate this Article