Update openwebtext_split.py

openwebtext_split.py

@@ -19,6 +19,7 @@ import os
 import re
 from itertools import chain
 import tarfile
+import lzma
 
 import datasets
 
@@ -55,11 +56,12 @@ def custom_iter_archive(path_or_buf, _filter=lambda x: True):
             if not file_path.endswith('xz'):
                 continue
             file_obj = stream.extractfile(tarinfo)
-            inner_stream = tarfile.open(file_obj=file_obj, mode="r|*")
-            for j, xzinfo in enumerate(inner_stream):
-                if not xzinfo.name.endswith('txt'):
-                    continue
-                yield xzinfo.name, inner_stream.extractfile(xzinfo)
+            with lzma.LZMAFile(file_obj) as inner_stream:
+                for j, xzinfo in enumerate(inner_stream):
+                    print(j, xzinfo)
+                    if not xzinfo.name.endswith('txt'):
+                        continue
+                    yield xzinfo.name, inner_stream.extractfile(xzinfo)
             stream.members = []
         del stream
     if hasattr(path_or_buf, "read"):