adds first dataloader

art.py

@@ -0,0 +1,84 @@
+import json
+import os
+
+import datasets
+
+
+_CITATION = """\
+@InProceedings{anli,
+  author = {Chandra, Bhagavatula and Ronan, Le Bras and Chaitanya, Malaviya and Keisuke, Sakaguchi and Ari, Holtzman
+    and Hannah, Rashkin and Doug, Downey and Scott, Wen-tau Yih and Yejin, Choi},
+  title = {Abductive Commonsense Reasoning},
+  year = {2020}
+}"""
+
+_DESCRIPTION = """\
+the Abductive Natural Language Generation Dataset from AI2
+"""
+_DATA_URL = "https://storage.googleapis.com/ai2-mosaic/public/abductive-commonsense-reasoning-iclr2020/anlg.zip"
+_HOMEPAGE = "https://github.com/allenai/abductive-commonsense-reasoning"
+
+class ArtConfig(datasets.BuilderConfig):
+    """BuilderConfig for Art."""
+
+    def __init__(self, **kwargs):
+        """BuilderConfig for Art.
+        Args:
+          **kwargs: keyword arguments forwarded to super.
+        """
+        super(ArtConfig, self).__init__(version=datasets.Version("0.1.0", ""), **kwargs)
+
+
+class Art(datasets.GeneratorBasedBuilder):
+    VERSION = datasets.Version("0.1.0")
+    BUILDER_CONFIGS = [
+        ArtConfig(
+            name="anlg",
+            description="""\
+            Abductive Natural Language Generation Dataset from AI2.
+            """,
+        ),
+    ]
+
+    def _info(self):
+        return datasets.DatasetInfo(
+            description=_DESCRIPTION,
+            features=datasets.Features(
+                {
+                    "gem_id": datasets.Value("string"),
+                    "observation_1": datasets.Value("string"),
+                    "observation_2": datasets.Value("string"),
+                    "label": datasets.Value("string"),
+                }
+            ),
+            homepage=_HOMEPAGE,
+            citation=_CITATION,
+        )
+
+    def _split_generators(self, dl_manager):
+        ds_splits = [datasets.Split.TRAIN, datasets.Split.VALIDATION, datasets.Split.TEST]
+        splits = ["train", "dev", "test"]
+        dl_dir = dl_manager.download_and_extract(_DATA_URL)
+
+        return [
+            datasets.SplitGenerator(
+                name=ds_split,
+                gen_kwargs={
+                    "filepath": os.path.join(dl_dir, "anlg", f"{split}-w-comet-preds.jsonl"),
+                    "split": split if split != "dev" else "validation" # adheres to GEM naming conventions
+                },
+            ) for ds_split, split in zip(ds_splits, splits)
+        ]
+
+    def _generate_examples(self, filepath, split):
+        with open(filepath, "r", encoding="utf-8") as f:
+            data = [json.loads(line) for line in f.readlines()]
+
+        for idx, row in enumerate(data):
+            label = row[f"hyp{row['label']}"]
+            yield idx, {
+                "gem_id": f"GEM-ART-{split}-{idx}",
+                "observation_1": row["obs1"],
+                "observation_2": row["obs2"],
+                "label": label,
+            }