Pass Variable Length Argument To Old Function Call

smdebug/tensorflow/utils.py

@@ -339,7 +339,7 @@ def get_layer_call_fn(layer: tf.keras.layers.Layer) -> Callable[[tf.Tensor], tf.
 
     def call(inputs, *args, **kwargs) -> tf.Tensor:
         layer_input = inputs
-        layer_output = old_call_fn(inputs)
+        layer_output = old_call_fn(inputs, *args, **kwargs)
         for hook in layer._hooks:
             hook_result = hook(inputs, layer_input=layer_input, layer_output=layer_output)
             if hook_result is not None:

tests/tensorflow2/test_model_subclassing.py

@@ -13,6 +13,16 @@ def __init__(self):
         self.conv1 = Conv2D(
             32, 3, activation="relu", kernel_initializer=tf.keras.initializers.GlorotNormal(seed=12)
         )
+        self.original_call = self.conv1.call
+
+        def new_call(inputs, *args, **kwargs):
+            # Since we use layer wrapper we need to assert if these parameters
+            # are actually being passed into the original call fn
+            assert kwargs["input_one"] == 1
+            kwargs.pop("input_one")
+            return self.original_call(inputs, *args, **kwargs)
+
+        self.conv1.call = new_call
         self.conv0 = Conv2D(
             32, 3, activation="relu", kernel_initializer=tf.keras.initializers.GlorotNormal(seed=12)
         )
@@ -26,7 +36,7 @@ def __init__(self):
     def first(self, x):
         with tf.name_scope("first"):
             tf.print("mymodel.first")
-            x = self.conv1(x)
+            x = self.conv1(x, input_one=1)
             # x = self.bn(x)
             return self.flatten(x)