Google I/O is a wrap! Catch up on TensorFlow sessions View sessions


Compute the split info on the given files.

Compute the split info (num shards, num examples,...) metadata required by tfds.folder_dataset.write_metadata.

See documentation for usage:

data_dir Directory containing the .tfrecord files (or similar format)
out_dir Output directory on which save the metadata. It should be available from the apache beam workers. If not set, apache beam won't be used (only available with some file formats).

split_infos The list of tfds.core.SplitInfo.