small bug fix for history return statement

src/run_modules/model_setup.py

@@ -63,11 +63,11 @@ class ModelSetup(RunEnvironment):
         self.data_store.set("model", self.model, self.scope)
 
     def _set_checkpoint(self):
+        """
+        Must be run after all callback functions that shall be tracked during training have been created (currently this
+        affects the learning rate decay and the advanced history [actually created in this method]).
+        """
         lr = self.data_store.get("lr_decay", scope="general.model")
-        # checkpoint = ModelCheckpoint(filepath=self.checkpoint_name, verbose=1, monitor='val_loss', save_best_only=True, mode='auto')
-        # checkpoint = ModelCheckpointAdvanced(filepath=self.checkpoint_name, verbose=1, monitor='val_loss',
-        #                                      save_best_only=True, mode='auto', callbacks_to_save=lr,
-        #                                      callbacks_filepath=self.callbacks_name)
         hist = HistoryAdvanced()
         self.data_store.set("hist", hist, scope="general.model")
         callbacks = [{"callback": lr, "path": self.callbacks_name % "lr"},

src/run_modules/training.py

@@ -89,7 +89,7 @@ class Training(RunEnvironment):
                                                # callbacks=self.callbacks)
                                                callbacks=[self.lr_sc, self.hist, self.checkpoint])
         else:
-            lr_filepath = self.checkpoint.callbacks[0]["path"]  # TODO: stopped here. why does training start 1 epoch too early or doesn't it?
+            lr_filepath = self.checkpoint.callbacks[0]["path"]
             hist_filepath = self.checkpoint.callbacks[1]["path"]
             lr_callbacks = pickle.load(open(lr_filepath, "rb"))
             hist_callbacks = pickle.load(open(hist_filepath, "rb"))
@@ -101,7 +101,7 @@ class Training(RunEnvironment):
                          {"callback": self.hist, "path": hist_filepath}]
             self.checkpoint.update_callbacks(callbacks)
             self.checkpoint.update_best(hist_callbacks)
-            self.hist = self.model.fit_generator(generator=self.train_set.distribute_on_batches(),
+            _ = self.model.fit_generator(generator=self.train_set.distribute_on_batches(),
                                                  steps_per_epoch=len(self.train_set),
                                                  epochs=self.epochs,
                                                  verbose=2,