Base HF dataset name on url of the croissant file

PiperOrigin-RevId: 640134142

Author: Alfonso Castaño

tensorflow_datasets/core/dataset_builders/croissant_builder.py

@@ -173,7 +173,9 @@ def __init__(
     if mapping is None:
       mapping = {}
     self.dataset = mlc.Dataset(jsonld, mapping=mapping)
-    self.name = huggingface_utils.convert_hf_name(self.dataset.metadata.name)
+    self.name = huggingface_utils.get_tfds_name_from_croissant_dataset(
+        self.dataset
+    )
     self.metadata = self.dataset.metadata
 
     # In TFDS, version is a mandatory attribute, while in Croissant it is only a

tensorflow_datasets/core/dataset_builders/huggingface_dataset_builder.py

@@ -203,7 +203,9 @@ def __init__(
     self._hf_repo_id = hf_repo_id
     self._hf_config = hf_config
     self.config_kwargs = config_kwargs
-    tfds_config = huggingface_utils.convert_hf_name(hf_config)
+    tfds_config = (
+        huggingface_utils.convert_hf_name(hf_config) if hf_config else None
+    )
     try:
       self._hf_builder = hf_datasets.load_dataset_builder(
           self._hf_repo_id, self._hf_config, **self.config_kwargs

tensorflow_datasets/core/utils/huggingface_utils.py

@@ -15,8 +15,11 @@
 
 """Utility functions for huggingface_dataset_builder."""
 
+from __future__ import annotations
+
 from collections.abc import Mapping, Sequence
 import datetime
+import typing
 from typing import Any, Type, TypeVar
 
 from etils import epath
@@ -30,6 +33,10 @@
 from tensorflow_datasets.core.utils.lazy_imports_utils import datasets as hf_datasets
 from tensorflow_datasets.core.utils.lazy_imports_utils import tensorflow as tf
 
+if typing.TYPE_CHECKING:
+  # pylint: disable=g-bad-import-order
+  import mlcroissant as mlc
+
 
 _HF_DTYPE_TO_NP_DTYPE = immutabledict.immutabledict({
     'bool': np.bool_,
@@ -242,7 +249,17 @@ def convert_hf_value(
   )
 
 
-def convert_hf_name(hf_name: _StrOrNone) -> _StrOrNone:
+def get_tfds_name_from_croissant_dataset(dataset: mlc.Dataset) -> str:
+  """Returns TFDS compatible dataset name of the given MLcroissant dataset."""
+  if (url := dataset.metadata.url) and url.startswith(
+      'https://huggingface.co/datasets/'
+  ):
+    url_suffix = url.removeprefix('https://huggingface.co/datasets/')
+    return convert_hf_name(url_suffix)
+  return convert_hf_name(dataset.metadata.name)
+
+
+def convert_hf_name(hf_name: str) -> str:
   """Converts Huggingface name to a TFDS compatible dataset name.
 
   Huggingface names can contain characters that are not supported in
@@ -259,8 +276,6 @@ def convert_hf_name(hf_name: _StrOrNone) -> _StrOrNone:
     The TFDS compatible dataset name (dataset names, config names and split
     names).
   """
-  if hf_name is None:
-    return hf_name
   hf_name = hf_name.lower().replace('/', '__')
   return py_utils.make_valid_name(hf_name)
 

tensorflow_datasets/core/utils/huggingface_utils_test.py

@@ -22,6 +22,7 @@
 from tensorflow_datasets.core import lazy_imports_lib
 from tensorflow_datasets.core import registered
 from tensorflow_datasets.core.utils import huggingface_utils
+from tensorflow_datasets.core.utils.lazy_imports_utils import mlcroissant as mlc
 from tensorflow_datasets.core.utils.lazy_imports_utils import tensorflow as tf
 
 
@@ -216,6 +217,29 @@ def test_convert_value(hf_value, feature, expected_value):
   assert huggingface_utils.convert_hf_value(hf_value, feature) == expected_value
 
 
+@pytest.mark.parametrize(
+    'croissant_name,croissant_url,tfds_name',
+    [
+        (
+            'Name+1',
+            'https://huggingface.co/datasets/HuggingFaceH4/ultrachat_200k',
+            'huggingfaceh4__ultrachat_200k',
+        ),
+        ('Name+1', 'bad_url', 'name_1'),
+        ('Name+1', None, 'name_1'),
+    ],
+)
+def test_get_tfds_name_from_croissant_dataset(
+    croissant_name, croissant_url, tfds_name
+):
+  metadata = mlc.Metadata(name=croissant_name, url=croissant_url)
+  dataset = mlc.Dataset.from_metadata(metadata)
+  assert (
+      huggingface_utils.get_tfds_name_from_croissant_dataset(dataset)
+      == tfds_name
+  )
+
+
 @pytest.mark.parametrize(
     'hf_name,tfds_name',
     [
@@ -228,7 +252,6 @@ def test_convert_value(hf_value, feature, expected_value):
         # Config and split names
         ('x.y', 'x_y'),
         ('x_v1.0', 'x_v1_0'),
-        (None, None),
     ],
 )
 def test_from_hf_to_tfds(hf_name, tfds_name):