losses introduced to server class

Author: gaborvecsei

fed_learn/fed_server.py

@@ -15,11 +15,15 @@ def __init__(self, model_fn: Callable,
         self.nb_clients = nb_clients
         self.weight_summarizer = weight_summarizer
 
+        # Initialize the global model's weights
         self.model_fn = model_fn
         model = self.model_fn()
         self.global_model_weights = model.get_weights()
         fed_learn.get_rid_of_the_models(model)
 
+        self.global_losses = []
+        self.epoch_losses = []
+
         (x_train, y_train), _ = datasets.cifar10.load_data()
 
         if only_debugging:
@@ -37,7 +41,6 @@ def __init__(self, model_fn: Callable,
 
         self.client_data_indices = None
         self.clients = []
-
         self.client_model_weights = []
 
         # Training parameters used by the clients
@@ -68,9 +71,11 @@ def send_model(self, client):
 
     def init_for_new_epoch(self):
         # Reset clients
-        self.clients = []
+        self.clients.clear()
         # Reset the collected weights
-        self.client_model_weights = []
+        self.client_model_weights.clear()
+        # Reset epoch losses
+        self.epoch_losses.clear()
         # Generate new data indices for the clients
         self._generate_data_indices()
 

fl.py

@@ -28,22 +28,23 @@ def model_fn():
     server.create_clients()
     server.init_for_new_epoch()
 
-    loss = []
-
     for client in server.clients:
         print("Client {0} is starting the training".format(client.id))
 
         server.send_model(client)
         server.send_train_data(client)
 
         hist = client.edge_train(server.get_client_train_param_dict())
-        loss.append(hist.history["loss"][-1])
+        server.epoch_losses.append(hist.history["loss"][-1])
 
         server.receive_results(client)
 
     server.summarize_weights()
-    print("Loss (mean): {0}".format(np.mean(loss)))
-    loss.clear()
+
+    epoch_mean_loss = np.mean(server.epoch_losses)
+    server.global_losses.append(epoch_mean_loss)
+    print("Loss (mean): {0}".format(server.global_losses[-1]))
+    
     print("-" * 30)
 
     # TODO: test the base model with the aggregated weights