custom data loader

hacdc-wikipedia.py

@@ -0,0 +1,179 @@
+import gzip
+import json
+import os
+import datasets
+from datasets import Features, Sequence, Value
+
+
+logger = datasets.logging.get_logger(__name__)
+
+
+_FILES = [
+    'kwnlp-enwiki-20220901-link-annotated-text1-p1p41242.jsonl.gz',
+    'kwnlp-enwiki-20220901-link-annotated-text2-p41243p151573.jsonl.gz',
+    'kwnlp-enwiki-20220901-link-annotated-text3-p151574p311329.jsonl.gz',
+    'kwnlp-enwiki-20220901-link-annotated-text4-p311330p558391.jsonl.gz',
+    'kwnlp-enwiki-20220901-link-annotated-text5-p558392p958045.jsonl.gz',
+    'kwnlp-enwiki-20220901-link-annotated-text6-p958046p1483661.jsonl.gz',
+    'kwnlp-enwiki-20220901-link-annotated-text7-p1483662p2134111.jsonl.gz',
+    'kwnlp-enwiki-20220901-link-annotated-text8-p2134112p2936260.jsonl.gz',
+    'kwnlp-enwiki-20220901-link-annotated-text9-p2936261p4045402.jsonl.gz',
+    'kwnlp-enwiki-20220901-link-annotated-text10-p4045403p5399366.jsonl.gz',
+    'kwnlp-enwiki-20220901-link-annotated-text11-p5399367p6899366.jsonl.gz',
+    'kwnlp-enwiki-20220901-link-annotated-text11-p6899367p7054859.jsonl.gz',
+    'kwnlp-enwiki-20220901-link-annotated-text12-p7054860p8554859.jsonl.gz',
+    'kwnlp-enwiki-20220901-link-annotated-text12-p8554860p9172788.jsonl.gz',
+    'kwnlp-enwiki-20220901-link-annotated-text13-p10672789p11659682.jsonl.gz',
+    'kwnlp-enwiki-20220901-link-annotated-text13-p9172789p10672788.jsonl.gz',
+    'kwnlp-enwiki-20220901-link-annotated-text14-p11659683p13159682.jsonl.gz',
+    'kwnlp-enwiki-20220901-link-annotated-text14-p13159683p14324602.jsonl.gz',
+    'kwnlp-enwiki-20220901-link-annotated-text15-p14324603p15824602.jsonl.gz',
+    'kwnlp-enwiki-20220901-link-annotated-text15-p15824603p17324602.jsonl.gz',
+    'kwnlp-enwiki-20220901-link-annotated-text15-p17324603p17460152.jsonl.gz',
+    'kwnlp-enwiki-20220901-link-annotated-text16-p17460153p18960152.jsonl.gz',
+    'kwnlp-enwiki-20220901-link-annotated-text16-p18960153p20460152.jsonl.gz',
+    'kwnlp-enwiki-20220901-link-annotated-text16-p20460153p20570392.jsonl.gz',
+    'kwnlp-enwiki-20220901-link-annotated-text17-p20570393p22070392.jsonl.gz',
+    'kwnlp-enwiki-20220901-link-annotated-text17-p22070393p23570392.jsonl.gz',
+    'kwnlp-enwiki-20220901-link-annotated-text17-p23570393p23716197.jsonl.gz',
+    'kwnlp-enwiki-20220901-link-annotated-text18-p23716198p25216197.jsonl.gz',
+    'kwnlp-enwiki-20220901-link-annotated-text18-p25216198p26716197.jsonl.gz',
+    'kwnlp-enwiki-20220901-link-annotated-text18-p26716198p27121850.jsonl.gz',
+    'kwnlp-enwiki-20220901-link-annotated-text19-p27121851p28621850.jsonl.gz',
+    'kwnlp-enwiki-20220901-link-annotated-text19-p28621851p30121850.jsonl.gz',
+    'kwnlp-enwiki-20220901-link-annotated-text19-p30121851p31308442.jsonl.gz',
+    'kwnlp-enwiki-20220901-link-annotated-text20-p31308443p32808442.jsonl.gz',
+    'kwnlp-enwiki-20220901-link-annotated-text20-p32808443p34308442.jsonl.gz',
+    'kwnlp-enwiki-20220901-link-annotated-text20-p34308443p35522432.jsonl.gz',
+    'kwnlp-enwiki-20220901-link-annotated-text21-p35522433p37022432.jsonl.gz',
+    'kwnlp-enwiki-20220901-link-annotated-text21-p37022433p38522432.jsonl.gz',
+    'kwnlp-enwiki-20220901-link-annotated-text21-p38522433p39996245.jsonl.gz',
+    'kwnlp-enwiki-20220901-link-annotated-text22-p39996246p41496245.jsonl.gz',
+    'kwnlp-enwiki-20220901-link-annotated-text22-p41496246p42996245.jsonl.gz',
+    'kwnlp-enwiki-20220901-link-annotated-text22-p42996246p44496245.jsonl.gz',
+    'kwnlp-enwiki-20220901-link-annotated-text22-p44496246p44788941.jsonl.gz',
+    'kwnlp-enwiki-20220901-link-annotated-text23-p44788942p46288941.jsonl.gz',
+    'kwnlp-enwiki-20220901-link-annotated-text23-p46288942p47788941.jsonl.gz',
+    'kwnlp-enwiki-20220901-link-annotated-text23-p47788942p49288941.jsonl.gz',
+    'kwnlp-enwiki-20220901-link-annotated-text23-p49288942p50564553.jsonl.gz',
+    'kwnlp-enwiki-20220901-link-annotated-text24-p50564554p52064553.jsonl.gz',
+    'kwnlp-enwiki-20220901-link-annotated-text24-p52064554p53564553.jsonl.gz',
+    'kwnlp-enwiki-20220901-link-annotated-text24-p53564554p55064553.jsonl.gz',
+    'kwnlp-enwiki-20220901-link-annotated-text24-p55064554p56564553.jsonl.gz',
+    'kwnlp-enwiki-20220901-link-annotated-text24-p56564554p57025655.jsonl.gz',
+    'kwnlp-enwiki-20220901-link-annotated-text25-p57025656p58525655.jsonl.gz',
+    'kwnlp-enwiki-20220901-link-annotated-text25-p58525656p60025655.jsonl.gz',
+    'kwnlp-enwiki-20220901-link-annotated-text25-p60025656p61525655.jsonl.gz',
+    'kwnlp-enwiki-20220901-link-annotated-text25-p61525656p62585850.jsonl.gz',
+    'kwnlp-enwiki-20220901-link-annotated-text26-p62585851p63975909.jsonl.gz',
+    'kwnlp-enwiki-20220901-link-annotated-text27-p63975910p65475909.jsonl.gz',
+    'kwnlp-enwiki-20220901-link-annotated-text27-p65475910p66975909.jsonl.gz',
+    'kwnlp-enwiki-20220901-link-annotated-text27-p66975910p68475909.jsonl.gz',
+    'kwnlp-enwiki-20220901-link-annotated-text27-p68475910p69975909.jsonl.gz',
+    'kwnlp-enwiki-20220901-link-annotated-text27-p69975910p71475909.jsonl.gz',
+    'kwnlp-enwiki-20220901-link-annotated-text27-p71475910p71655058.jsonl.gz',
+]
+
+
+_VERSION = datasets.Version("1.0.0")
+_DESCRIPTION = "Wikipedia text with metadata"
+
+
+FEATURES = {
+    "all": Features(
+        {
+            "page_id": Value("int64"),
+            "page_title": Value("string"),
+            "paragraphs": [
+                {
+                    "plaintext": Value("string"),
+                    "section_idx": Value("int64"),
+                    "section_name": Value("string"),
+                    "target_page_ids": Sequence(Value("int64")),
+                    "anchor_spans": Sequence(Sequence(Value("int64"), length=2)),
+                }
+            ],
+            "revision_id": Value("int64"),
+            "categories": Sequence(Value("string")),
+            "item_id": Value("int64"),
+            "in_link_count": Value("int64"),
+            "out_link_count": Value("int64"),
+            "tmpl_good_article": Value("bool"),
+            "tmpl_featured_article": Value("bool"),
+            "item_label": Value("string"),
+            "item_description": Value("string"),
+            "item_alias": Sequence(Value("string")),
+        }
+    ),
+    "text": Features(
+        {
+            "paragraph": Value("string")
+        }
+    ),
+    "text-intros": Features(
+        {
+            "paragraph": Value("string")
+        }
+    ),
+}
+
+
+class HacdcWikimedia20220901(datasets.GeneratorBasedBuilder):
+
+    BUILDER_CONFIGS = [
+        datasets.BuilderConfig(name="all", version=_VERSION, description=_DESCRIPTION),
+        datasets.BuilderConfig(name="text", version=_VERSION, description=_DESCRIPTION),
+        datasets.BuilderConfig(
+            name="text-intros", version=_VERSION, description=_DESCRIPTION
+        ),
+    ]
+
+    DEFAULT_CONFIG_NAME = "all"
+
+    def _info(self):
+        return datasets.DatasetInfo(
+            description=_DESCRIPTION,
+            features=FEATURES[self.config.name],
+        )
+
+
+    def _split_generators(self, dl_manager):
+
+        paths = [os.path.join("data", _FILES[0])]
+        downloaded_files = dl_manager.download(paths)
+        print("PATHS", paths)
+        print("DLFILES", downloaded_files)
+        return [
+            datasets.SplitGenerator(name="every", gen_kwargs={"filepaths": downloaded_files}),
+        ]
+
+
+    def _generate_examples(self, filepaths):
+        """Generate examples.
+        Args:
+          filepaths: a list of file paths
+        Yields:
+          dictionaries representing article data and metadata
+        """
+        _id = 0
+        for filepath in filepaths:
+            logger.info("generating examples from = %s", filepath)
+            print("FP", filepath)
+            with gzip.open(filepath, "r") as fp:
+                for line in fp:
+                    data = json.loads(line.decode("utf-8"))
+
+                    if self.config.name == "all":
+                        yield _id, data
+                        _id += 1
+                        continue
+
+                    for paragraph in data["paragraphs"]:
+                        if self.config.name == "text":
+                            yield _id, {"paragraph": paragraph["plaintext"]}
+                            _id += 1
+
+                        if self.config.name == "text-intros":
+                            if paragraph["section_idx"] == 0:
+                                yield _id, {"paragraph": paragraph["plaintext"]}
+                                _id += 1