Ignore codec errors

the_pile.py

@@ -237,7 +237,7 @@ class ThePile(datasets.GeneratorBasedBuilder):
                         else:
                             id_ = path.split("/")[-1].split(".")[0]
                             meta = {"id": id_}
-                        text = file.read().decode("utf-8")
+                        text = file.read().decode("utf-8", errors="ignore")  # encoding errors in enron_emails
                         yield key, {
                             "text": text,
                             "meta": meta,