Adjusted sparcs evaluation to confusion matrix output of predict function

sparcs/sparcs_04_eval.py

@@ -18,13 +18,13 @@ from sparcs.sparcs_02_dataset import (net, valid_ds, valid_dl, optimizer,
 
 if __name__ == '__main__':
 
-    # calculate accuracy for each batch in the validation set
-    accuracies = predict(net, valid_dl, optimizer, accuracy_function,
-                         state_file)
-    mean_acc = np.mean(accuracies)
-    print('After training for {:d} epochs, we achieved an overall mean '
-          'accuracy of {:.2f}%  on the validation set!'.format(net.epoch,
-                                                               mean_acc * 100))
+    # predict each batch in the validation set
+    cm = predict(net, valid_dl, optimizer, accuracy_function, state_file)
+
+    # calculate overal accuracy
+    acc = (cm.diag().sum() / cm.sum()).numpy().item()
+    print('After training for {:d} epochs, we achieved an overall accuracy of '
+          '{:.2f}%  on the validation set!'.format(net.epoch, acc * 100))
 
     # number of samples to plot
     n = 5