DEBUG:Divide by zero

config.ini

@@ -16,7 +16,7 @@ lr = 0.01
 [TRAIN_PARAMS]
 training_package = decentralizepy.training.Training
 training_class = Training
-epochs_per_round = 4
+epochs_per_round = 1
 batch_size = 1024
 shuffle = True
 loss_package = torch.nn

src/decentralizepy/datasets/Femnist.py

@@ -272,7 +272,7 @@ class Femnist(Dataset):
                     total_pred[label] += 1
                     total_predicted += 1
 
-        logging.debug("Predicted on the test set")
+        logging.info("Predicted on the test set")
 
         for key, value in enumerate(correct_pred):
             if total_pred[key] != 0:
@@ -283,7 +283,7 @@ class Femnist(Dataset):
 
         accuracy = 100 * float(total_correct) / total_predicted
         logging.info("Overall accuracy is: {:.1f} %".format(accuracy))
-        logging.debug("Evaluating complete.")
+        logging.info("Evaluating complete.")
 
 
 class LogisticRegression(nn.Module):

src/decentralizepy/node/Node.py

@@ -159,7 +159,9 @@ class Node:
             self.trainer.train(self.dataset)
 
             self.sharing.step()
-            self.optimizer = optimizer_class(self.model.parameters(), **optimizer_params) # Reset optimizer state
+            self.optimizer = optimizer_class(
+                self.model.parameters(), **optimizer_params
+            )  # Reset optimizer state
             self.trainer.reset_optimizer(self.optimizer)
 
             rounds_to_test -= 1

src/decentralizepy/training/Training.py

@@ -45,7 +45,6 @@ class Training:
 
     def reset_optimizer(self, optimizer):
         self.optimizer = optimizer
-        
 
     def train(self, dataset):
         """

testing.py

@@ -42,5 +42,5 @@ if __name__ == "__main__":
     mp.spawn(
         fn=Node,
         nprocs=procs_per_machine,
-        args=[m_id, l, g, my_config, 20, "results", logging.INFO],
+        args=[m_id, l, g, my_config, 20, "results", logging.DEBUG],
     )