Add a test for `tf.function` calls that return multiple datasets under the same parent op name.

This triggers a bug in the C++ runtime that fails to uniquely name the DatasetToGraphV2 JIT nodes.

PiperOrigin-RevId: 391626904

tensorflow_federated/python/core/impl/executors/executor_bindings_test.py

@@ -142,6 +142,39 @@ class TensorFlowExecutorBindingsTest(parameterized.TestCase,
                                 TensorType(sequence_type.element.dtype))
     self.assertEqual(result, sum(range(5)))
 
+  def test_create_tuple_of_value_sequence(self):
+    self.skipTest('b/197147669')
+    datasets = (tf.data.Dataset.range(5), tf.data.Dataset.range(5))
+    executor = executor_bindings.create_tensorflow_executor()
+    struct_of_sequence_type = StructType([
+        (None, SequenceType(datasets[0].element_spec)),
+        (None, SequenceType(datasets[0].element_spec))
+    ])
+    arg_value_pb, _ = value_serialization.serialize_value(
+        datasets, struct_of_sequence_type)
+    arg = executor.create_value(arg_value_pb)
+
+    @computations.tf_computation(struct_of_sequence_type)
+    def preprocess(datasets):
+
+      def double_value(x):
+        return 2 * x
+
+      @tf.function
+      def add_preprocessing(ds1, ds2):
+        return ds1.map(double_value), ds2.map(double_value)
+
+      return add_preprocessing(*datasets)
+
+    comp_pb = serialization_bindings.Value(
+        computation=preprocess.get_proto(preprocess))
+    comp = executor.create_value(comp_pb)
+    result = executor.create_call(comp.ref, arg.ref)
+    output_pb = executor.materialize(result.ref)
+    result, result_type_spec = value_serialization.deserialize_value(
+        output_pb, type_hint=struct_of_sequence_type)
+    self.assert_types_identical(result_type_spec, struct_of_sequence_type)
+
   def test_create_struct(self):
     executor = executor_bindings.create_tensorflow_executor()
     expected_type_spec = TensorType(shape=[3], dtype=tf.int64)