Convert dataset to Parquet

README.md

@@ -5,7 +5,20 @@ configs:
   data_files:
   - split: test
     path: summarization/test-*
+- config_name: qa_corrupted
+  data_files:
+  - split: test
+    path: qa_corrupted/test-*
+- config_name: qa_rewritten
+  data_files:
+  - split: test
+    path: qa_rewritten/test-*
+- config_name: summarization
+  data_files:
+  - split: test
+    path: summarization/test-*
 dataset_info:
+- config_name: default
   features:
   - name: source
     dtype: string
@@ -23,6 +36,64 @@ dataset_info:
     num_examples: 4200
   download_size: 3026042
   dataset_size: 5392937
+- config_name: qa_corrupted
+  features:
+  - name: source
+    dtype: string
+  - name: question
+    dtype: string
+  - name: answer
+    dtype: string
+  - name: passage
+    dtype: string
+  - name: entity
+    dtype: string
+  - name: entity_type
+    dtype: string
+  splits:
+  - name: test
+    num_bytes: 5505177
+    num_examples: 4200
+  download_size: 2471155
+  dataset_size: 5505177
+- config_name: qa_rewritten
+  features:
+  - name: source
+    dtype: string
+  - name: question
+    dtype: string
+  - name: answer
+    dtype: string
+  - name: passage
+    dtype: string
+  - name: entity
+    dtype: string
+  - name: entity_type
+    dtype: string
+  splits:
+  - name: test
+    num_bytes: 5091326
+    num_examples: 4200
+  download_size: 2417759
+  dataset_size: 5091326
+- config_name: summarization
+  features:
+  - name: source
+    dtype: string
+  - name: entity
+    dtype: string
+  - name: entity_type
+    dtype: string
+  - name: fact
+    dtype: string
+  - name: passage
+    dtype: string
+  splits:
+  - name: test
+    num_bytes: 5392937
+    num_examples: 4200
+  download_size: 2978637
+  dataset_size: 5392937
 ---
 
 Data for the SMeL Test benchmark (paper coming soon).

qa/rewritten/test-00000-of-00001.parquet

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:fb7d29a0c6c320cb7fc1c616b314fbeb7c6f23abd6034d9acb80f09ccfa1ac31
-size 2454983

{qa/corrupted → qa_corrupted}/test-00000-of-00001.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:289e92a2c9ce6999fc3b4290e5e6a6ba31cadd4e59591897526dd9ab8cc45bd5
-size 2508251
+oid sha256:874576add0f5f31c3a02f3661cfd90e65ebbdd2efb46f293b74ce58119eca778
+size 2471155

{qa/ground_truth → qa_rewritten}/test-00000-of-00001.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8c921f2d9e7a5b5f738f5d009666a4b72b1a7703b06fb331ac3617902df2db75
-size 341759
+oid sha256:4f359e6a47bd553d90d7aaf99f3dad2dc93ca2a3d784ecd9d517a1bde4dba01a
+size 2417759

smel.py

@@ -1,112 +0,0 @@
-import datasets
-from huggingface_hub import HfApi, HfFileSystem
-import pyarrow.parquet as pq
-
-_CITATION = """\
-@misc{gahdritz_smel_2025,
-  author={Gustaf Ahdritz},
-  title={SMEL Dataset},
-  year={2025},
-  publisher={Hugging Face Hub},
-  url={https://huggingface.co/datasets/gahdritz/smel}
-}
-"""
-
-_DESCRIPTION = "The SMeL Test"
-_HOMEPAGE = "https://huggingface.co/datasets/gahdritz/smel"
-_LICENSE = "cc_by_4.0"
-_HF_DATASET_REPO_ID = "gahdritz/smel"
-_BASE_URL_RESOLVE = f"https://huggingface.co/datasets/{_HF_DATASET_REPO_ID}/resolve/main/"
-
-_FEATURES_QA = datasets.Features({
-    "source": datasets.Value("string"),
-    "question": datasets.Value("string"),
-    "answer": datasets.Value("string"),
-    "passage": datasets.Value("string"),
-    "entity": datasets.Value("string"),
-    "entity_type": datasets.Value("string"),
-})
-
-_FEATURES_SUMMARIZATION = datasets.Features({
-    "source": datasets.Value("string"),
-    "entity": datasets.Value("string"),
-    "entity_type": datasets.Value("string"),
-    "fact": datasets.Value("string"),
-    "passage": datasets.Value("string"),
-})
-
-class SmelConfig(datasets.BuilderConfig):
-    def __init__(self, name, description, features, data_path, **kwargs):
-        super(SmelConfig, self).__init__()
-        self.name = name
-        self.description = description
-        self.features = features
-        self.data_path = data_path
-
-
-class SmelDataset(datasets.GeneratorBasedBuilder):
-    """SMeL dataset."""
-
-    BUILDER_CONFIGS = [
-        SmelConfig(
-            name="summarization",
-            description="Summarization tasks.",
-            features=_FEATURES_SUMMARIZATION,
-            data_path="summarization/",
-        ),
-        SmelConfig(
-            name="qa_rewritten",
-            description="Rewritten Question Answering tasks.",
-            features=_FEATURES_QA,
-            data_path="qa/rewritten",
-        ),
-        SmelConfig(
-            name="qa_corrupted",
-            description="Corrupted Question Answering tasks.",
-            features=_FEATURES_QA,
-            data_path="qa/corrupted",
-        ),
-    ]
-
-    def _info(self):
-        return datasets.DatasetInfo(
-            description=_DESCRIPTION,
-            features=self.config.features,
-            homepage=_HOMEPAGE,
-            license=_LICENSE,
-            citation=_CITATION,
-        )
-
-    def _split_generators(self, dl_manager):
-        """Returns SplitGenerators."""
-        fs = HfFileSystem() # Initialize the Hugging Face FileSystem
-        repo_path = f"datasets/{_HF_DATASET_REPO_ID}/" # Base path on the HF Filesystem
-        all_files_in_path = fs.ls(repo_path + self.config.data_path, detail=False)
-        parquet_files = [f for f in all_files_in_path if f.endswith('.parquet')]
-
-        data_urls = [
-             _BASE_URL_RESOLVE + f.replace(f"datasets/{_HF_DATASET_REPO_ID}/", "")
-             for f in parquet_files
-        ]
-
-        downloaded_files = dl_manager.download_and_extract(data_urls)
-
-        return [
-            datasets.SplitGenerator(
-                name=datasets.Split.TEST,
-                gen_kwargs={
-                    "filepaths": downloaded_files,
-                },
-            ),
-        ]
-  
-    def _generate_examples(self, filepaths):
-        """Yields examples."""
-        example_idx = 0
-        for filepath in filepaths:
-            table = pq.read_table(filepath)
-            pa_df = table.to_pandas()
-
-            for _, row in pa_df.iterrows():
-                yield example_idx, row.to_dict()
-                example_idx += 1

summarization/test-00000-of-00001.parquet

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:8ea71ec20064f63a1dc4f43e3c4bced929b87f9f052346b2b64e3ff137c9997d
-size 3026042