Fix TPUEmbedding layer to use named parameters

PiperOrigin-RevId: 384784083

tensorflow_recommenders/layers/embedding/tpu_embedding_layer.py

@@ -159,12 +159,17 @@ def _clone_and_prepare_features(feature_config):
     table_configs[config.table] = table_configs.get(
         config.table,
         tf.tpu.experimental.embedding.TableConfig(
-            config.table.vocabulary_size, config.table.dim,
-            config.table.initializer, config.table.optimizer,
-            config.table.combiner, config.table.name))
+            vocabulary_size=config.table.vocabulary_size,
+            dim=config.table.dim,
+            initializer=config.table.initializer,
+            optimizer=config.table.optimizer,
+            combiner=config.table.combiner,
+            name=config.table.name))
 
     output_objects.append(tf.tpu.experimental.embedding.FeatureConfig(
-        table_configs[config.table], config.max_sequence_length, config.name))
+        table=table_configs[config.table],
+        max_sequence_length=config.max_sequence_length,
+        name=config.name))
 
   # Fix up the optimizers.
   for _, new_table in table_configs.items():
@@ -206,8 +211,9 @@ def _update_table_configs(feature_config, table_config_map):
       raise ValueError("TableConfig %s does not match any of the TableConfigs "
                        "used to configure this layer." % config.table)
     output_objects.append(tf.tpu.experimental.embedding.FeatureConfig(
-        table_config_dict[config.table], config.max_sequence_length,
-        config.name))
+        table=table_config_dict[config.table],
+        max_sequence_length=config.max_sequence_length,
+        name=config.name))
 
   return tf.nest.pack_sequence_as(feature_config, output_objects)