Always wrap non-sequence types like maps in a list when tracing Model call

Avoids input_signature and unwrapping issues.

Fixes #26591 (again).

PiperOrigin-RevId: 238725001

tensorflow/python/keras/saving/saving_utils.py

@@ -18,6 +18,8 @@ from __future__ import absolute_import
 from __future__ import division
 from __future__ import print_function
 
+import collections
+
 from tensorflow.python.eager import def_function
 from tensorflow.python.framework import tensor_spec
 from tensorflow.python.util import nest
@@ -82,8 +84,13 @@ def trace_model_call(model, input_signature=None):
     input_specs = nest.pack_sequence_as(structure=inputs,
                                         flat_sequence=flat_input_specs)
     # The input signature of the call function is a list with one element, since
-    # all tensor inputs must be passed in as the first argument.
-    input_signature = [input_specs] if len(input_specs) > 1 else input_specs
+    # all tensor inputs must be passed in as the first argument. Single-element
+    # dictionaries and other non-sequence types must also be wrapped.
+    if (len(input_specs) > 1
+        or not isinstance(input_specs, collections.Sequence)):
+      input_signature = [input_specs]
+    else:
+      input_signature = input_specs
 
   # TODO(mdan): Should the model's call be autographed by default?
   @def_function.function(input_signature=input_signature, autograph=False)

tensorflow/python/keras/saving/saving_utils_test.py

@@ -28,12 +28,15 @@ from tensorflow.python import tf2
 from tensorflow.python.client import session as session_lib
 from tensorflow.python.eager import context
 from tensorflow.python.eager import def_function
+from tensorflow.python.feature_column import feature_column_v2
+from tensorflow.python.framework import constant_op
 from tensorflow.python.framework import dtypes
 from tensorflow.python.framework import ops
 from tensorflow.python.framework import tensor_spec
 from tensorflow.python.keras import backend as K
 from tensorflow.python.keras import keras_parameterized
 from tensorflow.python.keras import testing_utils
+from tensorflow.python.keras.engine import sequential
 from tensorflow.python.keras.saving import saving_utils
 from tensorflow.python.ops import array_ops
 from tensorflow.python.platform import test
@@ -130,6 +133,27 @@ class TraceModelCallTest(keras_parameterized.TestCase):
 
     self._assert_all_close(expected_outputs, signature_outputs)
 
+  @keras_parameterized.run_all_keras_modes
+  def test_trace_features_layer(self):
+    columns = [feature_column_v2.numeric_column('x')]
+    model = sequential.Sequential(
+        [feature_column_v2.DenseFeatures(columns)])
+    model_input = {'x': constant_op.constant([[1.]])}
+    model.predict(model_input, steps=1)
+    fn = saving_utils.trace_model_call(model)
+    self.assertAllClose({'output_1': [[1.]]}, fn({'x': [[1.]]}))
+
+    columns = [feature_column_v2.numeric_column('x'),
+               feature_column_v2.numeric_column('y')]
+    model = sequential.Sequential(
+        [feature_column_v2.DenseFeatures(columns)])
+    model_input = {'x': constant_op.constant([[1.]]),
+                   'y': constant_op.constant([[2.]])}
+    model.predict(model_input, steps=1)
+    fn = saving_utils.trace_model_call(model)
+    self.assertAllClose({'output_1': [[1., 2.]]},
+                        fn({'x': [[1.]], 'y': [[2.]]}))
+
   @keras_parameterized.run_all_keras_modes
   def test_specify_input_signature(self):
     model = testing_utils.get_small_sequential_mlp(10, 3, None)

tensorflow/python/saved_model/load_test.py

@@ -1186,6 +1186,18 @@ class LoadTest(test.TestCase, parameterized.TestCase):
         **model_input).values()
     self.assertAllClose([[1., 2.]], signature_output)
 
+  def test_dense_features_layer_fit(self, cycles):
+    columns = [feature_column_v2.numeric_column("x")]
+    model = sequential.Sequential(
+        [feature_column_v2.DenseFeatures(columns),
+         core.Dense(1)])
+    model_input = {"x": constant_op.constant([[1.]])}
+    model.compile(optimizer="adam", loss="mse")
+    model.fit(model_input, constant_op.constant([[3.]]))
+    loaded = self.cycle(model, cycles)
+    loaded._default_save_signature(model_input)
+    loaded.signatures["serving_default"](**model_input)
+
 
 class SingleCycleTests(test.TestCase, parameterized.TestCase):