Correctly measured the length of the input sequence: All inputs have same length now.

README.md

@@ -11,16 +11,16 @@ dataset_info:
     sequence: int64
   splits:
   - name: train
-    num_bytes: 1007949646
+    num_bytes: 590513519
     num_examples: 409035
   - name: validation
-    num_bytes: 73749943
+    num_bytes: 43203362
     num_examples: 30000
   - name: test
-    num_bytes: 98788672
+    num_bytes: 57865594
     num_examples: 40000
-  download_size: 82725960
-  dataset_size: 1180488261
+  download_size: 82209550
+  dataset_size: 691582475
 configs:
 - config_name: default
   data_files:

data/test-00000-of-00001.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:7c202d06893d46c307ce0dc94ff474aeb2f4a8d0baf524a6468261e6daee4452
-size 6918840
+oid sha256:1bd84954249f66e9ba9b29029f7e974d5be94affb537b2e8e1c5f2e92a1d9f9f
+size 6875317

data/{train-00000-of-00003.parquet → train-00000-of-00002.parquet}

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:519d6e2c43764965c1eac76badc3c1afd479efed9b88ecaea779d5863fcff794
-size 23583550
+oid sha256:a1b922bb1a43be21540eac73cf34aae7a42d120afd416933697ca4e534f0eb03
+size 35144846

data/{train-00001-of-00003.parquet → train-00001-of-00002.parquet}

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5b79d273cac2be0a159862117acc59b6b25bf78e18176ea16d1d0a0478bd94a5
-size 23544026
+oid sha256:ca5e69419d9fdf2eb1f717c47d24eeaac3fb3c6aeb38d87b0d0549c37df8c178
+size 35052274

data/train-00002-of-00003.parquet

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:af2b6e5dc84bfc1c3a18e7acc3cf7180cdb4bf7ecac93a2d40d6906353f5317f
-size 23510661

data/validation-00000-of-00001.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:df18844e5a744c802a1282838f0f96570040eed8f421eeaf3f7388a945f24b57
-size 5168883
+oid sha256:fb35069ea89e28178cccef8a1dcdfd2af08d09e1c54d3af42d06d17f2e654b78
+size 5137113