Fix save_weights when saving weights in TF format with a remote path.

`SupportWriteToRemote` was incorrectly guessing the file format and was assuming an H5 file, but the TF format was used and the folder failed to be copied.

PiperOrigin-RevId: 739001145

Author: hertschuh

tf_keras/saving/saving_api.py

@@ -24,6 +24,7 @@
 
 from tf_keras.saving import saving_lib
 from tf_keras.saving.legacy import save as legacy_sm_saving_lib
+from tf_keras.saving.legacy import saving_utils
 from tf_keras.utils import io_utils
 
 try:
@@ -75,8 +76,7 @@ class SupportWriteToRemote:
     supports remoted saved model out of the box.
     """
 
-    def __init__(self, filepath, overwrite=True, save_format=None):
-        save_format = get_save_format(filepath, save_format=save_format)
+    def __init__(self, filepath, overwrite, save_format):
         self.overwrite = overwrite
         if saving_lib.is_remote_path(filepath) and save_format != "tf":
             self.temp_directory = tempfile.TemporaryDirectory()
@@ -191,14 +191,14 @@ def save_model(model, filepath, overwrite=True, save_format=None, **kwargs):
     when loading the model. See the `custom_objects` argument in
     `tf.keras.saving.load_model`.
     """
+    save_format = get_save_format(filepath, save_format)
+
     # Supports remote paths via a temporary file
     with SupportWriteToRemote(
         filepath,
         overwrite=overwrite,
         save_format=save_format,
     ) as local_filepath:
-        save_format = get_save_format(filepath, save_format)
-
         # Deprecation warnings
         if save_format == "h5":
             warnings.warn(
@@ -307,8 +307,12 @@ def load_model(
 
 
 def save_weights(model, filepath, overwrite=True, **kwargs):
+    save_format = get_save_weights_format(filepath)
+
     # Supports remote paths via a temporary file
-    with SupportWriteToRemote(filepath, overwrite=overwrite) as local_filepath:
+    with SupportWriteToRemote(
+        filepath, overwrite=overwrite, save_format=save_format
+    ) as local_filepath:
         if str(local_filepath).endswith(".weights.h5"):
             # If file exists and should not be overwritten.
             try:
@@ -385,3 +389,12 @@ def get_save_format(filepath, save_format):
         return "tf"
     else:
         return "h5"
+
+
+def get_save_weights_format(filepath):
+    filepath = io_utils.path_to_string(filepath)
+    filepath_is_h5 = saving_utils.is_hdf5_filepath(filepath)
+    if filepath_is_h5:
+        return "h5"
+    else:
+        return "tf"

tf_keras/saving/saving_lib_test.py

@@ -533,26 +533,97 @@ def test_metadata(self):
         self.assertIn("keras_version", metadata)
         self.assertIn("date_saved", metadata)
 
-    def test_gfile_copy_called(self):
-        temp_filepath = Path(
-            os.path.join(self.get_temp_dir(), "my_model.keras")
+    def test_save_keras_gfile_copy_called(self):
+        path = Path(os.path.join(self.get_temp_dir(), "my_model.keras"))
+        model = keras.Sequential(
+            [
+                keras.Input(shape=(1, 1)),
+                keras.layers.Dense(4),
+            ]
         )
-        model = CompileOverridingModel()
         with mock.patch(
             "re.match", autospec=True
         ) as mock_re_match, mock.patch.object(
             tf.io.gfile, "copy"
         ) as mock_gfile_copy:
             # Check regex matching
             mock_re_match.return_value = True
-            model.save(temp_filepath, save_format="keras_v3")
+            model.save(path, save_format="keras_v3")
             mock_re_match.assert_called()
-            self.assertIn(str(temp_filepath), mock_re_match.call_args.args)
+            self.assertIn(str(path), mock_re_match.call_args.args)
 
             # Check gfile copied with filepath specified as destination
-            self.assertEqual(
-                str(temp_filepath), str(mock_gfile_copy.call_args.args[1])
-            )
+            mock_gfile_copy.assert_called()
+            self.assertEqual(str(path), str(mock_gfile_copy.call_args.args[1]))
+
+    def test_save_tf_gfile_copy_not_called(self):
+        path = Path(os.path.join(self.get_temp_dir(), "my_model.keras"))
+        model = keras.Sequential(
+            [
+                keras.Input(shape=(1, 1)),
+                keras.layers.Dense(4),
+            ]
+        )
+        with mock.patch(
+            "re.match", autospec=True
+        ) as mock_re_match, mock.patch.object(
+            tf.io.gfile, "copy"
+        ) as mock_gfile_copy:
+            # Check regex matching
+            mock_re_match.return_value = True
+            model.save(path, save_format="tf")
+            mock_re_match.assert_called()
+            self.assertIn(str(path), mock_re_match.call_args.args)
+
+            # Check gfile.copy was not used.
+            mock_gfile_copy.assert_not_called()
+
+    def test_save_weights_h5_gfile_copy_called(self):
+        path = Path(os.path.join(self.get_temp_dir(), "my_model.weights.h5"))
+        model = keras.Sequential(
+            [
+                keras.Input(shape=(1, 1)),
+                keras.layers.Dense(4),
+            ]
+        )
+        model(tf.constant([[1.0]]))
+        with mock.patch(
+            "re.match", autospec=True
+        ) as mock_re_match, mock.patch.object(
+            tf.io.gfile, "copy"
+        ) as mock_gfile_copy:
+            # Check regex matching
+            mock_re_match.return_value = True
+            model.save_weights(path)
+            mock_re_match.assert_called()
+            self.assertIn(str(path), mock_re_match.call_args.args)
+
+            # Check gfile copied with filepath specified as destination
+            mock_gfile_copy.assert_called()
+            self.assertEqual(str(path), str(mock_gfile_copy.call_args.args[1]))
+
+    def test_save_weights_tf_gfile_copy_not_called(self):
+        path = Path(os.path.join(self.get_temp_dir(), "my_model.ckpt"))
+        model = keras.Sequential(
+            [
+                keras.Input(shape=(1, 1)),
+                keras.layers.Dense(4),
+            ]
+        )
+        model(tf.constant([[1.0]]))
+        with mock.patch(
+            "re.match", autospec=True
+        ) as mock_re_match, mock.patch.object(
+            tf.io.gfile, "copy"
+        ) as mock_gfile_copy:
+            # Check regex matching
+            mock_re_match.return_value = True
+            model.save_weights(path)
+            mock_re_match.assert_called()
+            self.assertIn(str(path), mock_re_match.call_args.args)
+
+            # Check gfile.copy was not used.
+            mock_gfile_copy.assert_not_called()
 
     def test_load_model_api_endpoint(self):
         temp_filepath = Path(os.path.join(self.get_temp_dir(), "mymodel.keras"))