feat: Support custom target y column name for Bigframes Tensorflow

PiperOrigin-RevId: 592910297
googleapis · Dec 21, 2023 · 1634940 · 1634940
1 parent 6e6d005
commit 1634940
Show file tree

Hide file tree

Showing 3 changed files with 15 additions and 9 deletions.
diff --git a/tests/system/vertexai/test_bigframes_tensorflow.py b/tests/system/vertexai/test_bigframes_tensorflow.py
@@ -80,8 +80,7 @@ def test_remote_execution_keras(self, shared_state):
             "virginica": 1,
             "setosa": 2,
         }
-        df["target"] = df["species"].map(species_categories)
-        df = df.drop(columns=["species"])
+        df["species"] = df["species"].map(species_categories)
 
         train, _ = bf_train_test_split(df, test_size=0.2)
 
@@ -96,7 +95,10 @@ def test_remote_execution_keras(self, shared_state):
             enable_cuda=True,
             display_name=self._make_display_name("bigframes-keras-training"),
         )
-        model.fit.vertex.remote_config.serializer_args[train] = {"batch_size": 10}
+        model.fit.vertex.remote_config.serializer_args[train] = {
+            "batch_size": 10,
+            "target_col": "species",
+        }
 
         # Train model on Vertex
         model.fit(train, epochs=10)

diff --git a/tests/unit/vertexai/test_any_serializer.py b/tests/unit/vertexai/test_any_serializer.py
@@ -1106,6 +1106,7 @@ def test_any_serializer_deserialize_bigframe_tensorflow(
             any_serializer_instance._instances[serializers.BigframeSerializer],
             serialized_gcs_path=fake_gcs_path,
             batch_size=None,
+            target_col=None,
         )
 
     def test_any_serializer_deserialize_tf_dataset(

diff --git a/vertexai/preview/_workflow/serialization_engine/serializers.py b/vertexai/preview/_workflow/serialization_engine/serializers.py
@@ -1200,7 +1200,7 @@ def deserialize(
             return self._deserialize_torch(serialized_gcs_path)
         elif detected_framework == "tensorflow":
             return self._deserialize_tensorflow(
-                serialized_gcs_path, kwargs.get("batch_size")
+                serialized_gcs_path, kwargs.get("batch_size"), kwargs.get("target_col")
             )
         else:
             raise ValueError(f"Unsupported framework: {detected_framework}")
@@ -1273,14 +1273,18 @@ def reduce_tensors(a, b):
         return functools.reduce(reduce_tensors, list(parquet_df_dp))
 
     def _deserialize_tensorflow(
-        self, serialized_gcs_path: str, batch_size: Optional[int] = None
+        self,
+        serialized_gcs_path: str,
+        batch_size: Optional[int] = None,
+        target_col: Optional[str] = None,
     ) -> TFDataset:
         """Tensorflow deserializes parquet (GCS) --> tf.data.Dataset
 
         serialized_gcs_path is a folder containing one or more parquet files.
         """
-        # Set default batch_size
+        # Set default kwarg values
         batch_size = batch_size or DEFAULT_TENSORFLOW_BATCHSIZE
+        target_col = target_col.encode("ASCII") or b"target"
 
         # Deserialization at remote environment
         try:
@@ -1301,13 +1305,12 @@ def _deserialize_tensorflow(
             ds_shard = tfio.IODataset.from_parquet(file_name)
             ds = ds.concatenate(ds_shard)
 
-        # TODO(b/296474656) Parquet must have "target" column for y
         def map_fn(row):
-            target = row[b"target"]
+            target = row[target_col]
             row = {
                 k: tf.expand_dims(v, -1)
                 for k, v in row.items()
-                if k != b"target" and k != b"index"
+                if k != target_col and k != b"index"
             }
 
             def reduce_fn(a, b):