Fix: deserialize callbacks for fit params from dict definition (#1389)

* Fix: deserialize callbacks for fit params from dict definition

* Fix: formatting

* Fix: better checking for None callbacks

* Fix: update build_callbacks reference

gordo/machine/model/models.py

@@ -315,6 +315,10 @@ def get_params(self, **params):
         params = super().get_params(**params)
         params.update({"kind": self.kind})
         params.update(self.kwargs)
+        if self.kwargs.get("callbacks") is not None and any(
+            isinstance(callback, dict) for callback in self.kwargs["callbacks"]
+        ):
+            params["callbacks"] = serializer.build_callbacks(self.kwargs["callbacks"])
         return params
 
     def _prepare_model(self):

gordo/serializer/__init__.py

@@ -1,4 +1,8 @@
-from .from_definition import from_definition, load_params_from_definition
+from .from_definition import (
+    from_definition,
+    load_params_from_definition,
+    build_callbacks,
+)
 from .into_definition import into_definition, load_definition_from_params
 from .serializer import (
     dump,

gordo/serializer/from_definition.py

@@ -248,28 +248,6 @@ def _build_step(
         )
 
 
-def _build_callbacks(definitions: list):
-    """
-    Parameters
-    ----------
-    definitions
-        List of callbacks definitions
-
-    Examples
-    --------
-    >>> callbacks=_build_callbacks([{'tensorflow.keras.callbacks.EarlyStopping': {'monitor': 'val_loss,', 'patience': 10}}])
-    >>> type(callbacks[0])
-    <class 'keras.src.callbacks.early_stopping.EarlyStopping'>
-
-    Returns
-    -------
-    """
-    callbacks = []
-    for callback in definitions:
-        callbacks.append(_build_step(callback))
-    return callbacks
-
-
 def _load_param_classes(params: dict):
     """
     Inspect the params' values and determine if any can be loaded as a class.
@@ -350,7 +328,7 @@ def _load_param_classes(params: dict):
                     kwargs = _load_param_classes(sub_params)
                     params[key] = create_instance(Model, **kwargs)  # type: ignore
         elif key == "callbacks" and isinstance(value, list):
-            params[key] = _build_callbacks(value)
+            params[key] = build_callbacks(value)
     return params
 
 
@@ -367,3 +345,25 @@ def load_params_from_definition(definition: dict) -> dict:
             "Expected definition to be a dict," f"found: {type(definition)}"
         )
     return _load_param_classes(definition)
+
+
+def build_callbacks(definitions: list):
+    """
+    Parameters
+    ----------
+    definitions
+        List of callbacks definitions
+
+    Examples
+    --------
+    >>> callbacks=build_callbacks([{'tensorflow.keras.callbacks.EarlyStopping': {'monitor': 'val_loss,', 'patience': 10}}])
+    >>> type(callbacks[0])
+    <class 'keras.src.callbacks.early_stopping.EarlyStopping'>
+
+    Returns
+    -------
+    """
+    callbacks = []
+    for callback in definitions:
+        callbacks.append(_build_step(callback))
+    return callbacks

tests/gordo/machine/model/test_model.py

@@ -357,6 +357,8 @@ def test_keras_autoencoder_fits_callbacks():
     assert isinstance(first_callback, EarlyStopping)
     assert first_callback.monitor == "val_loss"
     assert first_callback.patience == 10
+    X, y = np.random.rand(10, 10), np.random.rand(10, 10)
+    model.fit(X, y)
 
 
 def test_parse_module_path():