Polyaxon was built with the following goals:

  • Modularity: The creation of a computation graph based on modular and understandable modules, with the possibility to reuse and share the module in subsequent usage.
  • Usability: Training a model should be easy enough, and should enable quick experimentations.
  • Configurable: Models and experiments could be created using a YAML/Json file, but also in python files.
  • Extensibility: The modularity and the extensive documentation of the code makes it easy to build and extend the set of provided modules.
  • Performance: Polyaxon is based on internal tensorflow code base and leverage the builtin distributed learning.
  • Data Preprocessing: Polyaxon provides many pipelines and data processor to support different data inputs.


This work is based and was inspired from different projects, tensorflow.contrib.learn, keras, sonnet, and seq2seq. The idea behind creating this library is to provide a tool that allow engineers and researchers to develop and experiment with end to end solutions.

The choice of creating a new library was very important to have a complete control over the apis and future design decisions.