From 57bb5eb58d1f6ef6424d7166e5036aee38c1ff7e Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Adri=C3=A0=20Arrufat?=
 <1671644+arrufat@users.noreply.github.com>
Date: Tue, 31 Mar 2020 09:20:50 +0900
Subject: [PATCH] use running stats to track losses (#2041)

---
 examples/dnn_dcgan_train_ex.cpp | 13 ++++++++-----
 1 file changed, 8 insertions(+), 5 deletions(-)

diff --git a/examples/dnn_dcgan_train_ex.cpp b/examples/dnn_dcgan_train_ex.cpp
index 8eca83bdd..e9fb87d88 100644
--- a/examples/dnn_dcgan_train_ex.cpp
+++ b/examples/dnn_dcgan_train_ex.cpp
@@ -181,6 +181,7 @@ int main(int argc, char** argv) try
     const std::vector<float> fake_labels(minibatch_size, -1);
     dlib::image_window win;
     resizable_tensor real_samples_tensor, fake_samples_tensor, noises_tensor;
+    running_stats<double> g_loss, d_loss;
     while (iteration < 50000)
     {
         // Train the discriminator with real images
@@ -192,7 +193,7 @@ int main(int argc, char** argv) try
         }
         // The following lines are equivalent to calling train_one_step(real_samples, real_labels)
         discriminator.to_tensor(real_samples.begin(), real_samples.end(), real_samples_tensor);
-        double d_loss = discriminator.compute_loss(real_samples_tensor, real_labels.begin());
+        d_loss.add(discriminator.compute_loss(real_samples_tensor, real_labels.begin()));
         discriminator.back_propagate_error(real_samples_tensor);
         discriminator.update_parameters(d_solvers, learning_rate);
 
@@ -210,7 +211,7 @@ int main(int argc, char** argv) try
         // 4. finally train the discriminator and wait for the threading to stop.  The following
         // lines are equivalent to calling train_one_step(fake_samples, fake_labels)
         discriminator.to_tensor(fake_samples.begin(), fake_samples.end(), fake_samples_tensor);
-        d_loss += discriminator.compute_loss(fake_samples_tensor, fake_labels.begin());
+        d_loss.add(discriminator.compute_loss(fake_samples_tensor, fake_labels.begin()));
         discriminator.back_propagate_error(fake_samples_tensor);
         discriminator.update_parameters(d_solvers, learning_rate);
 
@@ -223,7 +224,7 @@ int main(int argc, char** argv) try
         // seen as test_one_step() plus the error back propagation.
 
         // Forward the fake samples and compute the loss with real labels
-        const auto g_loss = discriminator.compute_loss(fake_samples_tensor, real_labels.begin());
+        g_loss.add(discriminator.compute_loss(fake_samples_tensor, real_labels.begin()));
         // Back propagate the error to fill the final data gradient
         discriminator.back_propagate_error(fake_samples_tensor);
         // Get the gradient that will tell the generator how to update itself
@@ -238,10 +239,12 @@ int main(int argc, char** argv) try
             serialize("dcgan_sync") << generator << discriminator << iteration;
             std::cout <<
                 "step#: " << iteration <<
-                "\tdiscriminator loss: " << d_loss <<
-                "\tgenerator loss: " << g_loss << '\n';
+                "\tdiscriminator loss: " << d_loss.mean() * 2 <<
+                "\tgenerator loss: " << g_loss.mean() << '\n';
             win.set_image(tile_images(fake_samples));
             win.set_title("DCGAN step#: " + to_string(iteration));
+            d_loss.clear();
+            g_loss.clear();
         }
     }