Ultra-fast dataset upload with pre-encoded bytes

README.md

@@ -106,10 +106,10 @@ dataset_info:
     num_bytes: 750652393
     num_examples: 16685
   - name: val
-    num_bytes: 60955753
-    num_examples: 1355
-  download_size: 11279389272
-  dataset_size: 1997507652
+    num_bytes: 107692935
+    num_examples: 2362
+  download_size: 11358300575
+  dataset_size: 2044244834
 configs:
 - config_name: comparison
   data_files:

streaming/val-00000-of-00002.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e58d4e7d747545cd7e10320085c2151ecca30e9f6574790092a3682d36c73a96
-size 22276726
+oid sha256:0cb49abe1fa5d96d87832f216a64944f44c633f3bcc96915a3036aebc8db9bfa
+size 39272933

streaming/val-00001-of-00002.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:7601d31ca0e643a7fb0579058515fac53b6ce8b2e6dd58c2ddf3a8804b74231a
-size 22448387
+oid sha256:0971cec30556b7764fde91e168456f1c1f0076b7a7f66953b04a721b69bb856e
+size 39638370