simplify sort_values call

NVIDIA-Merlin · Apr 17, 2024 · e22a836 · e22a836
1 parent 700b196
commit e22a836
Showing 1 changed file with 1 addition and 7 deletions.
diff --git a/nvtabular/ops/categorify.py b/nvtabular/ops/categorify.py
@@ -1251,13 +1251,7 @@ def _drop_first_row(part, index):
             if has_size:
                 # Avoid using dask_cudf to calculate divisions
                 # (since it may produce too-few partitions)
-                df = df.sort_values(
-                    name_size,
-                    ascending=False,
-                    divisions=dd.shuffle._calculate_divisions(
-                        df, df[name_size], False, df.npartitions
-                    )[0][::-1],
-                )
+                df = df.sort_values(name_size, ascending=False)
 
             unique_path = _save_encodings(
                 df,