add sample code

RedPajama-Data-V2.py

@@ -36,7 +36,7 @@ with open("_QUALITY_SIGNAL_TAGS", "r") as f:
 
 _URL_BASE = 'https://data.together.xyz/redpajama-data-v2/v1.0.0'
 _LANGUAGES = ("en", "de", "fr", "es", "it")
-_DATA_DIR = os.environ.get('RED_PAJAMA_V2_DATA_DIR', None)
+_SAMPLE_SNAPSHOT_ID = "2023-06"
 
 _LISTINGS_PATTERN = "urls/{language}-{snapshot}-{partition}.txt"
 
@@ -55,7 +55,25 @@ class RedPajamaDataV2Config(datasets.BuilderConfig):
         self.language = language
 
 
-_BUILDER_CONFIGS = []
+_BUILDER_CONFIGS = [
+    RedPajamaDataV2Config(
+        name=f'_sample',
+        partition='sample',
+        snapshots=None,
+        language=None,
+        version=datasets.Version("1.0.0", ""),
+        description=f"RedPajamaV2 Sample",
+    ),
+    # this one is just an alias for the sample
+    RedPajamaDataV2Config(
+        name=f'sample',
+        partition='sample',
+        snapshots=None,
+        language=None,
+        version=datasets.Version("1.0.0", ""),
+        description=f"RedPajamaV2 Sample",
+    )
+]
 
 for lang in _LANGUAGES:
     _BUILDER_CONFIGS.extend(
@@ -141,7 +159,41 @@ class RedPajamaV2(datasets.GeneratorBasedBuilder):
                 supervised_keys=None,
             )
 
-    def _split_generators(self, dl_manager):
+    def _split_generators_sample(self, dl_manager):
+        # fetch documents
+        with open("sample/sample_listings.txt", "r") as fd:
+            listings = [line.strip() for line in fd]
+
+        # fetch documents
+        docs_files = dl_manager.download({
+            _SAMPLE_SNAPSHOT_ID: [
+                f"sample/documents/{lst}.json.gz" for lst in listings
+            ]
+        })
+
+        # fetch quality signals
+        signals_files = dl_manager.download({
+            _SAMPLE_SNAPSHOT_ID: [
+                f"sample/quality_signals/{lst}.signals.json.gz"
+                for lst in listings
+            ]
+        })
+
+        return [
+            datasets.SplitGenerator(
+                name=datasets.Split.TRAIN,
+                gen_kwargs={
+                    "documents_files": {
+                        _SAMPLE_SNAPSHOT_ID: docs_files[_SAMPLE_SNAPSHOT_ID]
+                    },
+                    "quality_signals_files": {
+                        _SAMPLE_SNAPSHOT_ID: signals_files[_SAMPLE_SNAPSHOT_ID]
+                    }
+                }
+            )
+        ]
+
+    def _split_generators_full(self, dl_manager):
         url_lists = dl_manager.download_and_extract({
             snapshot_id: _LISTINGS_PATTERN.format(
                 language=self.config.language,
@@ -157,40 +209,35 @@ class RedPajamaV2(datasets.GeneratorBasedBuilder):
             with open(listings_file, encoding="utf-8") as f:
                 listings_ids[snapshot_id] = [line.strip() for line in f]
 
-        # mapping document type -> url for download
-        if _DATA_DIR is not None:
-            documents_files = None
-            quality_signals_files = None
-        else:
-            # build urls pointing to documents
-            document_urls = {
+        # build urls pointing to documents
+        document_urls = {
+            snapshot_id: [
+                os.path.join(_URL_BASE, f"documents/{lst_id}.json.gz")
+                for lst_id in listings_ids[snapshot_id]
+            ]
+            for snapshot_id in self.config.snapshots
+        }
+
+        documents_files = dl_manager.download(document_urls)
+
+        # build urls pointing to quality signals
+        if self.config.partition == "head_middle":
+            quality_signals_urls = {
                 snapshot_id: [
-                    os.path.join(_URL_BASE, f"documents/{lst_id}.json.gz")
+                    os.path.join(
+                        _URL_BASE,
+                        f"quality_signals/{lst_id}.signals.json.gz"
+                    )
                     for lst_id in listings_ids[snapshot_id]
                 ]
                 for snapshot_id in self.config.snapshots
             }
 
-            documents_files = dl_manager.download(document_urls)
-
-            # build urls pointing to quality signals
-            if self.config.partition == "head_middle":
-                quality_signals_urls = {
-                    snapshot_id: [
-                        os.path.join(
-                            _URL_BASE,
-                            f"quality_signals/{lst_id}.signals.json.gz"
-                        )
-                        for lst_id in listings_ids[snapshot_id]
-                    ]
-                    for snapshot_id in self.config.snapshots
-                }
-
-                quality_signals_files = dl_manager.download(
-                    quality_signals_urls
-                )
-            else:
-                quality_signals_files = {}
+            quality_signals_files = dl_manager.download(
+                quality_signals_urls
+            )
+        else:
+            quality_signals_files = {}
 
         return [
             datasets.SplitGenerator(
@@ -208,6 +255,12 @@ class RedPajamaV2(datasets.GeneratorBasedBuilder):
             )
         ]
 
+    def _split_generators(self, dl_manager):
+        if self.config.name == "sample":
+            return self._split_generators_sample(dl_manager)
+
+        return self._split_generators_full(dl_manager)
+
     def _generate_examples(self, documents_files, quality_signals_files):
         """ This function returns examples """
         snapshots = list(documents_files.keys())
@@ -215,7 +268,7 @@ class RedPajamaV2(datasets.GeneratorBasedBuilder):
         key = 0
         for snapshot in snapshots:
             docs_files = documents_files[snapshot]
-            if self.config.partition == "head_middle":
+            if self.config.partition in ("head_middle", "sample"):
                 qs_files = quality_signals_files[snapshot]
             else:
                 qs_files = None
@@ -234,6 +287,7 @@ class RedPajamaV2(datasets.GeneratorBasedBuilder):
 
                                 meta = {
                                     "url": doc["url"],
+                                    "language": doc["language"],
                                     "source_domain": doc["source_domain"],
                                     "date_download": doc["date_download"],
                                     "digest": doc["digest"],

sample/sample_listings.txt

@@ -0,0 +1,100 @@
+documents/2023-06/0003/es_middle
+documents/2023-06/0003/de_head
+documents/2023-06/0003/fr_head
+documents/2023-06/0003/es_head
+documents/2023-06/0003/fr_middle
+documents/2023-06/0003/en_head
+documents/2023-06/0003/it_middle
+documents/2023-06/0003/en_middle
+documents/2023-06/0003/it_head
+documents/2023-06/0003/de_middle
+documents/2023-06/0004/es_middle
+documents/2023-06/0004/de_head
+documents/2023-06/0004/fr_head
+documents/2023-06/0004/es_head
+documents/2023-06/0004/fr_middle
+documents/2023-06/0004/en_head
+documents/2023-06/0004/it_middle
+documents/2023-06/0004/en_middle
+documents/2023-06/0004/it_head
+documents/2023-06/0004/de_middle
+documents/2023-06/0005/es_middle
+documents/2023-06/0005/de_head
+documents/2023-06/0005/fr_head
+documents/2023-06/0005/es_head
+documents/2023-06/0005/fr_middle
+documents/2023-06/0005/en_head
+documents/2023-06/0005/it_middle
+documents/2023-06/0005/en_middle
+documents/2023-06/0005/it_head
+documents/2023-06/0005/de_middle
+documents/2023-06/0002/es_middle
+documents/2023-06/0002/de_head
+documents/2023-06/0002/fr_head
+documents/2023-06/0002/es_head
+documents/2023-06/0002/fr_middle
+documents/2023-06/0002/en_head
+documents/2023-06/0002/it_middle
+documents/2023-06/0002/en_middle
+documents/2023-06/0002/it_head
+documents/2023-06/0002/de_middle
+documents/2023-06/0007/es_middle
+documents/2023-06/0007/de_head
+documents/2023-06/0007/fr_head
+documents/2023-06/0007/es_head
+documents/2023-06/0007/fr_middle
+documents/2023-06/0007/en_head
+documents/2023-06/0007/it_middle
+documents/2023-06/0007/en_middle
+documents/2023-06/0007/it_head
+documents/2023-06/0007/de_middle
+documents/2023-06/0000/es_middle
+documents/2023-06/0000/de_head
+documents/2023-06/0000/fr_head
+documents/2023-06/0000/es_head
+documents/2023-06/0000/fr_middle
+documents/2023-06/0000/en_head
+documents/2023-06/0000/it_middle
+documents/2023-06/0000/en_middle
+documents/2023-06/0000/it_head
+documents/2023-06/0000/de_middle
+documents/2023-06/0009/es_middle
+documents/2023-06/0009/de_head
+documents/2023-06/0009/fr_head
+documents/2023-06/0009/es_head
+documents/2023-06/0009/fr_middle
+documents/2023-06/0009/en_head
+documents/2023-06/0009/it_middle
+documents/2023-06/0009/en_middle
+documents/2023-06/0009/it_head
+documents/2023-06/0009/de_middle
+documents/2023-06/0008/es_middle
+documents/2023-06/0008/de_head
+documents/2023-06/0008/fr_head
+documents/2023-06/0008/es_head
+documents/2023-06/0008/fr_middle
+documents/2023-06/0008/en_head
+documents/2023-06/0008/it_middle
+documents/2023-06/0008/en_middle
+documents/2023-06/0008/it_head
+documents/2023-06/0008/de_middle
+documents/2023-06/0001/es_middle
+documents/2023-06/0001/de_head
+documents/2023-06/0001/fr_head
+documents/2023-06/0001/es_head
+documents/2023-06/0001/fr_middle
+documents/2023-06/0001/en_head
+documents/2023-06/0001/it_middle
+documents/2023-06/0001/en_middle
+documents/2023-06/0001/it_head
+documents/2023-06/0001/de_middle
+documents/2023-06/0006/es_middle
+documents/2023-06/0006/de_head
+documents/2023-06/0006/fr_head
+documents/2023-06/0006/es_head
+documents/2023-06/0006/fr_middle
+documents/2023-06/0006/en_head
+documents/2023-06/0006/it_middle
+documents/2023-06/0006/en_middle
+documents/2023-06/0006/it_head
+documents/2023-06/0006/de_middle