feat(loader): added dataset loader

baneks.py

@@ -0,0 +1,65 @@
+from pandas import read_csv
+
+from datasets import GeneratorBasedBuilder, Value, Version, BuilderConfig, Features, DatasetInfo, SplitGenerator, Split
+
+_DESCRIPTION = '''
+This dataset contains anekdotes parsed from a few vk social network communities. The data can be useful for fine-tuning language generation models as well for tasks of automatic humour analysis.
+'''
+
+_HOMEPAGE = 'https://huggingface.co/datasets/zeio/baneks'
+
+_LICENSE = 'Apache License Version 2.0'
+
+_URLS = {
+    'censored': 'https://huggingface.co/datasets/zeio/baneks/resolve/main/censored.tsv',
+    'default': 'https://huggingface.co/datasets/zeio/baneks/resolve/main/default.tsv',
+    'inflated': 'https://huggingface.co/datasets/zeio/baneks/resolve/main/inflated.tsv'
+}
+
+
+class Baneks(GeneratorBasedBuilder):
+
+    VERSION = Version('10.10.2023')
+
+    BUILDER_CONFIGS = [
+        BuilderConfig(name = 'censored', version = VERSION, description = 'No duplicates - entries with the same text are grouped and aggregated'),
+        BuilderConfig(name = 'default', version = VERSION, description = 'Same as "censored", but censored words are replaced with inferred values for their initial form'),
+        BuilderConfig(name = 'inflated', version = VERSION, description = 'Each entry corresponds to a post, minimal changes to the source data')
+    ]
+
+    DEFAULT_CONFIG_NAME = 'default'
+
+    def _info(self):
+        return DatasetInfo(
+            description=_DESCRIPTION,
+            features = Features({
+                'text': Value('string'),
+                'published': Value('string'),
+                'id': Value('int32'),
+                'n-likes': Value('int32'),
+                'n-views': Value('int32'),
+                'accessed': Value('string'),
+                'source': Value('string')
+            }),
+            homepage=_HOMEPAGE,
+            license=_LICENSE
+        )
+
+    def _split_generators(self, dl_manager):
+        name = self.config.name
+
+        url = _URLS[name]
+        # path = os.path.join(dl_manager.download_and_extract(url), f'{name}.tsv')
+
+        return [
+            SplitGenerator(
+                name = Split.TRAIN,
+                gen_kwargs = {
+                    "path": dl_manager.download_and_extract(url)
+                }
+            )
+        ]
+
+    def _generate_examples(self, path: str):
+        for _, row in read_csv(path, sep = '\t').iterrows():
+            yield f'{row["id"]:08d}-{row["source"]}', dict(row)