v0.7.5

Metadata can now be saved with the model with sofa.tl.save_model()
Please note that there are some exception where metadata can not be saved with the model as h5mu format:

• Too many columns
• Mixed type columns

In these cases, the model is saved without the metadata. The metadata then needs to be saved and loaded manually by the user (e.g. as a .csv file)

v0.7.4

The factors and loadings dataframes returned by sofa.tl.get_factors() and sofa.tl.get_loadings() have now the correct indices.

0.7.3

Changed Factor indexing to 1 based instead of 0. E.g. if you want to specify Factor 1, index is now 1 instead of 0.

Affects every function that needs to specify a Factor index:

• sofa.tl.get_top_loadings()
• sofa.pl.plot_top_loadings()
• sofa.pl.plot_loadings()

v0.7.2

changed function names:

• get_Z changed to get_factors
• get_W changed to get_loadings

fixed bug in sofa.tl.load_model

v0.7.1

Fixed bug in sofa.tl.save_model():
Device is now saved as str, because torch.device can't be saved to disk as MuData.
If old models have model.device set as torch.device it is now converted to str.