Solaris is a python toolbox for running tensorflow and pytorch machine learning pipelines for geospatial data.


Solaris is available in Puhti with following versions:

  • 0.1.3 (with tensorflow 1.13.1 and pytorch 1.1.0)


Solaris is included in the solaris module and can be loaded with

module load solaris

Solaris module includes a conda environment that has the Solaris library and many additional geospatial and machine learning libraries


For this version of Solaris, you should choose to use pytorch on your scripts as the included tensorflow can't use Puhti GPU at the moment due to incompatibility issues

Additional Python libraries in the solaris module

  • pandas (0.24.2)
  • geopandas (0.5.0)
  • rasterio (1.0.24)
  • scikit-image (0.15.0)
  • gdal (2.4.1)
  • networkx (2.3)
  • tensorflow (1.13.1)
  • pytorch (1.1.0)
  • And many more, for retrieving the full list in Puhti use: list-packages

License and citing

Solaris toolset is licensed with Apache Licence 2.0. Read the full license here.

In your publications please acknowledge also oGIIR and CSC, for example “The authors wish to acknowledge for computational resources CSC – IT Center for Science, Finland (urn:nbn:fi:research-infras-2016072531) and the Open Geospatial Information Infrastructure for Research (oGIIR, urn:nbn:fi:research-infras-2016072513).”