Fix keras model initialization after restart (#1602)

* Address case when keras model has not been evaluated and returns empty metrics

Signed-off-by: Patrick Foley <psfoley@gmail.com>

* Address review comments

Signed-off-by: Patrick Foley <psfoley@gmail.com>

* Fix lint formatting issue

Signed-off-by: Patrick Foley <psfoley@gmail.com>

---------

Signed-off-by: Patrick Foley <psfoley@gmail.com>

Author: psfoley

openfl/federated/task/runner_keras.py

@@ -169,6 +169,11 @@ def train_task(
 
         return global_tensor_dict, local_tensor_dict
 
+    def _initialize_metrics_result(self, batch_size):
+        # evaluation needed before metrics can be resolved
+        self.model.evaluate(self.data_loader.get_valid_loader(batch_size), verbose=1)
+        return self.model.get_metrics_result()
+
     def train_(self, batch_generator, metrics: list = None, **kwargs):
         """Train single epoch. Override this function for custom training.
 
@@ -187,15 +192,14 @@ def train_(self, batch_generator, metrics: list = None, **kwargs):
         #  initialization (build_model).
         #  If metrics are added (i.e. not a subset of what was originally
         #  defined) then the model must be recompiled.
+
+        batch_size = kwargs.get("batch_size", 1)
         try:
             results = self.model.get_metrics_result()
+            if len(results) == 0:
+                results = self._initialize_metrics_result(batch_size)
         except ValueError:
-            if "batch_size" in kwargs:
-                batch_size = kwargs["batch_size"]
-            else:
-                batch_size = 1
-            # evaluation needed before metrics can be resolved
-            self.model.evaluate(self.data_loader.get_valid_loader(batch_size), verbose=1)
+            self._initialize_metrics_result(batch_size)
             results = self.model.get_metrics_result()
 
         # TODO if there are new metrics in the flplan that were not included
@@ -230,10 +234,7 @@ def validate_task(self, col_name, round_num, input_tensor_dict, **kwargs):
                 These correspond to acc, precision, f1_score, etc.
             dict: Empty dictionary.
         """
-        if "batch_size" in kwargs:
-            batch_size = kwargs["batch_size"]
-        else:
-            batch_size = 1
+        batch_size = kwargs.get("batch_size", 1)
 
         self.rebuild_model(round_num, input_tensor_dict, validation=True)
         param_metrics = kwargs["metrics"]