Commit e227519
Use ShardBasedBuilder for HuggingFace datasets
This is a more efficient way and simpler way to generate HuggingFace datasets. This can save a lot of time and space, especially for large datasets.
This works on Beam and non-Beam.
PiperOrigin-RevId: 6822250151 parent 26812e9 commit e227519
2 files changed
Lines changed: 104 additions & 217 deletions
File tree
- tensorflow_datasets/core/dataset_builders
0 commit comments