Skip to content

Commit

Permalink
Fixes hash_crossed with cudf 21.12 (#1376)
Browse files Browse the repository at this point in the history
  • Loading branch information
Alberto Alvarez authored Jan 31, 2022
1 parent 4120980 commit f1f51b1
Show file tree
Hide file tree
Showing 3 changed files with 7 additions and 8 deletions.
3 changes: 1 addition & 2 deletions nvtabular/ops/hashed_cross.py
Original file line number Diff line number Diff line change
Expand Up @@ -59,8 +59,7 @@ def transform(self, col_selector: ColumnSelector, df: DataFrameType) -> DataFram
for cross in _nest_columns(col_selector.names):
val = 0
for column in cross:
val ^= _hash_series(df[column]) # or however we want to do this aggregation

val = _hash_series(df[column]) ^ val # or however we want to do this aggregation
if isinstance(self.num_buckets, dict):
val = val % self.num_buckets[cross]
else:
Expand Down
6 changes: 3 additions & 3 deletions requirements-gpu.txt
Original file line number Diff line number Diff line change
@@ -1,5 +1,5 @@
-r requirements.txt
cudf>=21.08
dask-cudf>=21.08
dask-cuda>=21.08
cudf>=21.12
dask-cudf>=21.12
dask-cuda>=21.12
cupy>=7
6 changes: 3 additions & 3 deletions requirements.txt
Original file line number Diff line number Diff line change
@@ -1,7 +1,7 @@
dask==2021.09.1
distributed==2021.09.1
dask==2021.11.2
distributed==2021.11.2
pandas>=1.2.0
numba>=0.54.0
numba>=0.55.1
tqdm>=4.0
pyarrow>=1.0
tensorflow-metadata>=1.2.0
Expand Down

0 comments on commit f1f51b1

Please sign in to comment.