Welcome to gpvisc’s documentation!

Copyright (2024) C. Le Losq and co.

gpvisc is a machine learning model trained to predict the viscosity of phospho-alumino-silicate melts.

It is trained on a extensive database, comprising more than 5,000 different melt compositions for a total of more than 28,000 viscosity data points. For some compositions like peridotite, predictions are even possible at pressures up to 30 GPa. The database is available here. The code is open source on Github.com/charlesll/gpvisc.

To use the model, we provide gpvisc as a Python library. Jupyter Notebooks are provided as examples of use. A Streamlit version is under construction.

Please follow the tutorials of this documentation to perform predictions!

If you use this package, please cite it using this citation key:

Le Losq C., Ferraina C., Sossi P. A., Boukaré C.-É. (2024). gpvisc. Zenodo. https://doi.org/10.5281/zenodo.13843250

Contents:

• Introduction
  • The database
  • The GP model
• Installation
• Inputs
  • Input composition for GP
  • Indicate a composition in a Python dictionnary
  • Import a composition from Excel/Libre Office
  • Dealing with iron oxidation state
  • Final preparation of melt composition
  • Temperature and pressure
  • Helper functions to create queries
  • Final preparation for import in GP model
• Predictions
  • Making a prediction
  • Prediction outputs
  • Checking for extrapolation
• Tutorials
  • Simple prediction
  • Peridotite
  • Redox
  • Benchmark on SciGlass (full Database)
  • Check extrapolations
  • Benchmark speed
  • Comparison to Giordano et al. 2008
• Web service
• Authors
  • List of contributors
  • Author contribution statement
• References

Indices and tables

• Index

• Module Index

• Search Page