Load only arabic language in c4/multilingual about datasets

[fylux]

For C4 (https://www.tensorflow.org/datasets/catalog/c4), "ar" is not a config but a split over the multilingual config so your second command is almost correct. Actually, if you run it you get the following error:

TypeError: load() takes 1 positional argument but 2 were given

In the documentation of tfds.load:
https://www.tensorflow.org/datasets/api_docs/python/tfds/load

tfds.load(
    name: str,
    *,
    split: Optional[Tree[splits_lib.SplitArg]] = None,
    ...

The star "*" after name means that the rest of arguments need to be specified by parameter name. So the correct command would be:

ds = tfds.load('c4/multilingual', split='ar')

Hope that helps! Still, keep in mind that (as described in https://www.tensorflow.org/datasets/catalog/c4), preparing this dataset requires some additional work compared with other datasets.