.

README.md

@@ -5,3 +5,10 @@ A dataset of all the sentences in Wikipedia.
 Filtered to only include sentences <=64 characters.
 
 Taken from the OPTIMUS project. https://github.com/ChunyuanLI/Optimus/blob/master/download_datasets.md
+
+The dataset is 11.8GB so best to load it using streaming:
+
+```python
+from datasets import load_dataset
+dataset = load_dataset("Fraser/wiki_sentences", split='train', streaming=True)
+```

wiki_sentences.py

@@ -39,6 +39,8 @@ class WikiSentences(datasets.GeneratorBasedBuilder):
     def _generate_examples(self, filepath):
         with open(filepath, encoding="utf-8") as txt_file:
             for i, line in enumerate(txt_file):
+                import pdb
+                pdb.set_trace()
                 line = line.strip()
-                if len(line) <= 64:
+                if line and len(line) <= 64:
                     yield i, {"text": line}