diff --git a/keras/backend/common/variables.py b/keras/backend/common/variables.py
index 8eee4752c2c..24e0320c5fb 100644
--- a/keras/backend/common/variables.py
+++ b/keras/backend/common/variables.py
@@ -48,7 +48,7 @@ def __init__(
             if callable(initializer):
                 self._value = None
                 self._initializer = initializer
-                self._shape = standardize_shape(shape)
+                self._shape = self._validate_shape(shape)
                 register_uninitialized_variable(self)
             else:
                 raise ValueError(
@@ -70,6 +70,7 @@ def __init__(
                 )
         else:
             if callable(initializer):
+                shape = self._validate_shape(shape)
                 value = initializer(shape, dtype=dtype)
             else:
                 value = initializer
@@ -91,6 +92,16 @@ def _deferred_initialize(self):
         value = self._initializer(self._shape, dtype=self._dtype)
         self._initialize(value)
 
+    def _validate_shape(self, shape):
+        shape = standardize_shape(shape)
+        if None in shape:
+            raise ValueError(
+                "Shapes used to initialize variables must be "
+                "fully-defined (no `None` dimensions). Received: "
+                f"shape={shape} for variable path='{self.path}'"
+            )
+        return shape
+
     def _maybe_autocast(self, value):
         autocast_scope = get_autocast_scope()
         if autocast_scope is not None:
diff --git a/keras/metrics/f_score_metrics.py b/keras/metrics/f_score_metrics.py
index 2bac6d44372..4a03afffa87 100644
--- a/keras/metrics/f_score_metrics.py
+++ b/keras/metrics/f_score_metrics.py
@@ -135,7 +135,7 @@ def _build(self, y_true_shape, y_pred_shape):
             )
         num_classes = y_pred_shape[-1]
         if self.average != "micro":
-            init_shape = num_classes
+            init_shape = (num_classes,)
         else:
             init_shape = ()