In streaming mode shuffle train files

mc4_nl_cleaned.py

@@ -19,6 +19,7 @@ import json
 import gzip
 import textwrap
 import datasets
+import random
 from itertools import zip_longest
 
 logger = datasets.logging.get_logger(__name__)
@@ -151,6 +152,9 @@ class Mc4(datasets.GeneratorBasedBuilder):
                             n_shards=8 if split == "validation" else 1024,
                         )
                     )
+        # Shuffle data in streaming mode, so restarts will not always start with the same data
+        if dl_manager.is_streaming:
+            random.shuffle(data_urls["train"])
         train_downloaded_files = dl_manager.download(data_urls["train"])
         validation_downloaded_files = dl_manager.download(data_urls["validation"])
         return [

test_mc4_nl_cleaned.py

@@ -13,9 +13,9 @@ def test_streaming_dataset():
 
     i = iter(train_ds)
     e = next(i)
-    assert len(e["text"]) == 1264
+    assert len(e["text"]) != 1264
     e = next(i)
-    assert len(e["text"]) == 1091
+    assert len(e["text"]) != 1091
 
     assert sum(1 for _ in iter(val_ds)) == 16189