Saving various data types to disk #22
Description
tree_config.yaml package is meant to be dealing with precisely this issue. I.e. to be able to dump anything as a string with yaml_dumps that can then be loaded again as needed. And it's meant to handle numpy/torch and other basic types that h5 may not be able to represent by default.
Or you can try a different approach e.g. just calling repr(value). But for config stuff I think tree_config should work as that is how things are saved/loaded already. So instead of creating a dataset for each item, I'd create a group and save all the properties as metadata dict values with str being both key and value.
Originally posted by @matham in #1 (comment)