szha closed pull request #9805: Enable the reporting of cross-entropy or nll 
loss value when training CNN network using the models defined by 

This is a PR merged from a forked repository.
As GitHub hides the original diff on merge, it is displayed below for
the sake of provenance:

As this is a foreign pull request (from a fork), the diff is supplied
below (as it won't show otherwise due to GitHub magic):

diff --git a/example/image-classification/common/ 
index d9f96d0eba..0e0cd521f2 100755
--- a/example/image-classification/common/
+++ b/example/image-classification/common/
@@ -117,6 +117,8 @@ def add_fit_args(parser):
                        help='load the model on an epoch using the 
     train.add_argument('--top-k', type=int, default=0,
                        help='report the top-k accuracy. 0 means no report.')
+    train.add_argument('--loss', type=str, default='',
+                       help='show the cross-entropy or nll loss. ce strands 
for cross-entropy, nll-loss stands for likelihood loss')
     train.add_argument('--test-io', type=int, default=0,
                        help='1 means test reading speed without training')
     train.add_argument('--dtype', type=str, default='float32',
@@ -260,6 +262,23 @@ def fit(args, network, data_loader, **kwargs):
             'top_k_accuracy', top_k=args.top_k))
+    supported_loss = ['ce', 'nll_loss']
+    if len(args.loss) > 0:
+        # ce or nll loss is only applicable to softmax output
+        loss_type_list = args.loss.split(',')
+        if 'softmax_output' in network.list_outputs():
+            for loss_type in loss_type_list:
+                loss_type = loss_type.strip()
+                if loss_type == 'nll':
+                    loss_type = 'nll_loss'
+                if loss_type not in supported_loss:
+                    logging.warning(loss_type + ' is not an valid loss type, 
only cross-entropy or ' \
+                                    'negative likelihood loss is supported!')
+                else:
+                    eval_metrics.append(mx.metric.create(loss_type))
+        else:
+            logging.warning("The output is not softmax_output, loss argument 
will be skipped!")
     # callbacks that run after each batch
     batch_end_callbacks = [mx.callback.Speedometer(
         args.batch_size, args.disp_batches)]


This is an automated message from the Apache Git Service.
To respond to the message, please log on GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:

With regards,
Apache Git Services

Reply via email to