v2.2.0

Full Changelog: v2.1.4...v2.2.0

• All properties can now be computed using only a single forward pass in the neural network using the model.predict_all() function. This limits the computing time and is much more efficient and thus frugal, particularly when doing MC Dropout. The old API still can be used, but it is encouraged to use predict_all(), which returns a dictionary that contains all the latent variables and observed properties.
• the predict() function API has changed: several properties can now be asked passing a list to the function, for the argument methods. The old argument method still works too. Therefore, when interest in several properties (latent or observed), please use the new methodsargument. It will do only one forward pass in the neural network to make all the predictions. This results in better efficiency and significantly less computing time when making multiple query.
• Code has been cleaned and some functions not needed for i-Melt were removed.
• the Streamlit app has been updated and is now much much faster, thanks to the new predict()behavior.
• Training a neural network is also faster as we can group viscosity predictions: we gained a factor 2 approximately.
• All example notebooks have been updated accordingly.

v2.1.4

Update minimal version of Python to 3.10 as it was creating problems with older release and dependencies

Full Changelog: v2.1.3...v2.1.4

v2.1.3

Full Changelog: v2.1.2...v2.1.3

v2.1.2

Full Changelog: v2.1.1...v2.1.2

v2.1.1

What's Changed

See NEWS.md

Full Changelog: v2.1.0...v2.1.1

Merge pull request #2 from charlesll/v2.1.0

Update the code to make i-Melt a PyPI package.

Update of results notebook

Update of results notebook following paper review

i-melt v2.0.0

This is the release 2.0.0 of i-Melt, which now allows prediction of melt/glass properties in the CaO-MgO-Na2O-K2O-Al2O3-SiO2 system.

i-Melt v1.2.1

This release is associated with the accepted publication in Geochimica Cosmochimica Acta (https://www.sciencedirect.com/science/article/abs/pii/S0016703721005007). This version of i-Melt allows prediction of melt/glass structural and thermodynamic properties for alkali aluminosilicate compositions. This sub-release includes changes in ReadMe.

i-Melt v1.2: alkali aluminosilicate compositions

This release is associated with the accepted publication in Geochimica Cosmochimica Acta (https://www.sciencedirect.com/science/article/abs/pii/S0016703721005007). This version of i-Melt allows prediction of melt/glass structural and thermodynamic properties for alkali aluminosilicate compositions.