Merge branch 'main' into add-grouped-natural-id-partitioner

datasets/doc/source/how-to-use-with-local-data.rst

@@ -37,14 +37,6 @@ CSV
   data_files = [ "path-to-my-file-1.csv", "path-to-my-file-2.csv", ...]
   dataset = load_dataset("csv", data_files=data_files)
 
-  # Divided Dataset
-  data_files = {
-    "train": single_train_file_or_list_of_files,
-    "test": single_test_file_or_list_of_files,
-    "can-have-more-splits": ...
-  }
-  dataset = load_dataset("csv", data_files=data_files)
-
   partitioner = ChosenPartitioner(...)
   partitioner.dataset = dataset
   partition = partitioner.load_partition(partition_id=0)
@@ -60,18 +52,10 @@ JSON
   # Single file
   data_files = "path-to-my-file.json"
 
-  # Multitple Files
+  # Multiple Files
   data_files = [ "path-to-my-file-1.json", "path-to-my-file-2.json", ...]
   dataset = load_dataset("json", data_files=data_files)
 
-  # Divided Dataset
-  data_files = {
-    "train": single_train_file_or_list_of_files,
-     "test": single_test_file_or_list_of_files,
-     "can-have-more-splits": ...
-  }
-  dataset = load_dataset("json", data_files=data_files)
-
   partitioner = ChosenPartitioner(...)
   partitioner.dataset = dataset
   partition = partitioner.load_partition(partition_id=0)
@@ -103,7 +87,12 @@ Then, the path you can give is `./mnist`.
   from flwr_datasets.partitioner import ChosenPartitioner
 
   # Directly from a directory
-  dataset = load_dataset("imagefolder", data_dir="/path/to/folder")
+  dataset_dict = load_dataset("imagefolder", data_dir="/path/to/folder")
+  # Note that what we just loaded is a DatasetDict, we need to choose a single split
+  # and assign it to the partitioner.dataset
+  # e.g. "train" split but that depends on the structure of your directory
+  dataset = dataset_dict["train"]
+
   partitioner = ChosenPartitioner(...)
   partitioner.dataset = dataset
   partition = partitioner.load_partition(partition_id=0)
@@ -134,7 +123,11 @@ Analogously to the image datasets, there are two methods here:
   from datasets import load_dataset
   from flwr_datasets.partitioner import ChosenPartitioner
 
-  dataset = load_dataset("audiofolder", data_dir="/path/to/folder")
+  dataset_dict = load_dataset("audiofolder", data_dir="/path/to/folder")
+  # Note that what we just loaded is a DatasetDict, we need to choose a single split
+  # and assign it to the partitioner.dataset
+  # e.g. "train" split but that depends on the structure of your directory
+  dataset = dataset_dict["train"]
 
   partitioner = ChosenPartitioner(...)
   partitioner.dataset = dataset
@@ -230,7 +223,7 @@ Partitioner abstraction is designed to allow for a single dataset assignment.
 
 .. code-block:: python
 
-  partitioner.dataset = your_dataset
+  partitioner.dataset = your_dataset # (your_dataset must be of type dataset.Dataset)
 
 If you need to do the same partitioning on a different dataset, create a new Partitioner
 for that, e.g.:

datasets/flwr_datasets/partitioner/partitioner.py

@@ -50,6 +50,11 @@ def dataset(self, value: Dataset) -> None:
                 "created partitions (in case the partitioning scheme needs to create "
                 "the full partitioning also in order to return a single partition)."
             )
+        if not isinstance(value, Dataset):
+            raise TypeError(
+                f"The dataset object you want to assign to the partitioner should be "
+                f"of type `datasets.Dataset` but given {type(value)}."
+            )
         self._dataset = value
 
     @abstractmethod

datasets/flwr_datasets/partitioner/partitioner_test.py

@@ -0,0 +1,59 @@
+# Copyright 2024 Flower Labs GmbH. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+# ==============================================================================
+"""Abstract partitioner tests."""
+
+
+import unittest
+
+import datasets
+from datasets import Dataset
+from flwr_datasets.partitioner.partitioner import Partitioner
+
+
+class DummyPartitioner(Partitioner):
+    """Dummy partitioner for testing."""
+
+    def load_partition(self, partition_id: int) -> Dataset:
+        """Return always a dummy dataset."""
+        return datasets.Dataset.from_dict({"feature": [0, 1, 2]})
+
+    @property
+    def num_partitions(self) -> int:
+        """Return always 0."""
+        return 0
+
+
+class TestPartitioner(unittest.TestCase):
+    """Test Partitioner."""
+
+    def test_dataset_setter_incorrect_type(self) -> None:
+        """Test if the incorrect type of the dataset to dataset.setter method raises."""
+        train_split = datasets.Dataset.from_dict({"feature": [0, 1, 2]})
+        test_split = datasets.Dataset.from_dict({"feature": [0, 1, 2]})
+        dataset = datasets.DatasetDict({"train": train_split, "test": test_split})
+        partitioner = DummyPartitioner()
+
+        with self.assertRaises(Exception) as context:
+            partitioner.dataset = dataset
+        self.assertIn(
+            "The dataset object you want to assign to the partitioner should be of "
+            "type `datasets.Dataset` but given "
+            "<class 'datasets.dataset_dict.DatasetDict'>.",
+            str(context.exception),
+        )
+
+
+if __name__ == "__main__":
+    unittest.main()