|View source on GitHub|
Returns the subsplit of the data for the process.
tfds.split_for_jax_process( split: str, *, process_index:
tfds.typing.Dim= None, process_count:
tfds.typing.Dim= None, drop_remainder: bool = False ) ->
In distributed setting, all process/hosts should get a non-overlapping, equally sized slice of the entire data. This function takes as input a split and extracts the slice for the current process index.
This funtion is an alias for:
By default, if examples can't be evenly distributed across processes, you can
drop extra examples with
The sub-split of the given