indexing through parquet file for faster streaming?

dataset.py

@@ -12,7 +12,8 @@ class ForestSegmentationDataset(datasets.GeneratorBasedBuilder):
         )
 
     def _split_generators(self, dl_manager):
-        sample_stream = dl_manager.iter_jsonl("index.jsonl")
+        #sample_stream = dl_manager.iter_jsonl("index.jsonl")
+        sample_stream = dl_manager.iter_jsonl("index.parquet")
         return [
             datasets.SplitGenerator(
                 name=datasets.Split.TRAIN,

index.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:48cf36efdaf1c1428e4656ed7bb06a1a17c522813fda069f512a41bde1ff65d1
+size 1245372