Fix testing errors cause by recent TF updates.

- Remove exact check for metric_embedding config dict. This class is a
  thin wrapper around the layers.Dense class and doesn't take any custom
  args.
- Update the deserialize from identifier functions to raise the
  ValueError if we are not able to cover the identifier to the target
  object type.

tensorflow_similarity/distances/__init__.py

@@ -110,12 +110,13 @@ def get(identifier) -> Distance:
     Raises:
         ValueError: If `identifier` cannot be interpreted.
     """
-    if isinstance(identifier, Distance):
-        return identifier
-    elif isinstance(identifier, dict):
-        return deserialize(identifier)
+    if isinstance(identifier, dict):
+        identifier = deserialize(identifier)
     elif isinstance(identifier, str):
         config = {"class_name": str(identifier), "config": {}}
-        return deserialize(config)
+        identifier = deserialize(config)
+
+    if isinstance(identifier, Distance):
+        return identifier
     else:
         raise ValueError("Could not interpret search identifier: {}".format(identifier))

tensorflow_similarity/search/__init__.py

@@ -117,12 +117,13 @@ def get(identifier, **kwargs) -> Search:
     Raises:
         ValueError: If `identifier` cannot be interpreted.
     """
-    if isinstance(identifier, Search):
-        return identifier
-    elif isinstance(identifier, dict):
-        return deserialize(identifier)
+    if isinstance(identifier, dict):
+        identifier = deserialize(identifier)
     elif isinstance(identifier, str):
         config = {"class_name": str(identifier), "config": kwargs}
-        return deserialize(config)
+        identifier = deserialize(config)
+
+    if isinstance(identifier, Search):
+        return identifier
     else:
         raise ValueError("Could not interpret search identifier: {}".format(identifier))

tensorflow_similarity/stores/__init__.py

@@ -109,12 +109,13 @@ def get(identifier) -> Store:
     Raises:
         ValueError: If `identifier` cannot be interpreted.
     """
-    if isinstance(identifier, Store):
-        return identifier
-    elif isinstance(identifier, dict):
-        return deserialize(identifier)
+    if isinstance(identifier, dict):
+        identifier = deserialize(identifier)
     elif isinstance(identifier, str):
         config = {"class_name": str(identifier), "config": {}}
-        return deserialize(config)
+        identifier = deserialize(config)
+
+    if isinstance(identifier, Store):
+        return identifier
     else:
         raise ValueError("Could not interpret Store identifier: {}".format(identifier))

tests/test_layers.py

@@ -1,6 +1,7 @@
 import math
 
 import tensorflow as tf
+from tensorflow.keras import layers
 
 from tensorflow_similarity.layers import (
     GeneralizedMeanPooling1D,
@@ -160,29 +161,6 @@ def test_metric_embedding(self):
         expected_result = tf.constant([[0.5, 0.5, 0.5, 0.5], [0.5, 0.5, 0.5, 0.5]])
         self.assertAllClose(result, expected_result, rtol=1e-06)
 
-    def test_metric_embedding_get_config(self):
-        me_layer = MetricEmbedding(32)
-        config = me_layer.get_config()
-        expected_config = {
-            "name": "metric_embedding",
-            "trainable": True,
-            "dtype": "float32",
-            "units": 32,
-            "activation": "linear",
-            "use_bias": True,
-            "kernel_initializer": {
-                "class_name": "GlorotUniform",
-                "config": {"seed": None},
-            },
-            "bias_initializer": {"class_name": "Zeros", "config": {}},
-            "kernel_regularizer": None,
-            "bias_regularizer": None,
-            "activity_regularizer": None,
-            "kernel_constraint": None,
-            "bias_constraint": None,
-        }
-        self.assertEqual(expected_config, config)
-
 
 if __name__ == "__main__":
     tf.test.main()