Version 1.4.0

Python environment migration

Hopsworks back end mechanism for managing projects’ Python environment is now more scalable and robust by moving to Docker based Anaconda environments. That means Python environments of existing projects will default back to the base Python environment provided by Hopsworks, so all additional Python libraries that were installed will be lost.

Users can find a YAML file that contains the exported Anaconda environment of the project including any user-installed libraries. This file can be found in the Resources dataset with a name python_env_<project>.yml where <project> is the project name. Users can use this YAML file to refer to previously installed Python libraries and then manually install them through the UI or the REST API. Importing the exported environment is discouraged as the base environment has been modified from the previous version and certain libraries such as TensorFlow will malfunction.

TensorFlow 2.x support

As of this version, Hopsworks projects are shipped by default with TensorFlow 2. TensorFlow 1.x is no longer supported automatically configured into the python dependencies of projects..