Dataset now reads all the tfrecord files

wikipedia_bert_128.py

@@ -142,8 +142,7 @@ class WikipediaBERT128(datasets.GeneratorBasedBuilder):
             'masked_lm_ids',        # masked_lm_labels=None      : label of masked tokens with padding as 0.
             'next_sentence_labels'  # next_sentence_label=None   : 1 if next sentence, 0 otherwise
         )
-        # tfrecords = Path(filepath).glob("*.tfrecord")
-        tfrecords = [Path(filepath)/"wiki_000.tfrecord"]
+        tfrecords = Path(filepath).glob("*.tfrecord")
 
         highest_id_ = -1
         for rec in tfrecords: