Create a script called ae_build_hdf5

[fdumais]

It should work with a toy dataset, maybe we should add it to Zenodo ?

[jhlegarreta]

maybe we should add it to Zenodo ?

Not sure. Usually one puts something in zenodo in order to share and be able to cite it, and I do not think a testing dataset is destined to be cited.

Other possibilities:

• Add a data folder in tractolearn with lightweight data. Would serve to test the rest of the scripts as well (assuming we fetch the weights from zenodo and the testing data is one that can be assimilated to the trained model). pyproject.toml might need to be adapted so that such data is available when installing the package.
• Host the data elsewhere (e.g. another GitHub repository, OSF, etc.). That would entail adding new entries to the fetcher, and probably requiring to create a separate enum so that the current one (targeted to "production" purposes) is not polluted.