fix zero tensor (#719)

* fix 0 tensor

education/homework2.md

@@ -182,7 +182,7 @@ for pass_id in range(100):
         optimizerD.clear_grad()
 
         errD = errD_real + errD_fake
-        losses[0].append(errD.numpy()[0])
+        losses[0].append(float(errD))
 
         ############################
         # (2) Update G network: maximize log(D(G(z)))
@@ -197,7 +197,7 @@ for pass_id in range(100):
         optimizerG.step()
         optimizerG.clear_grad()
 
-        losses[1].append(errG.numpy()[0])
+        losses[1].append(float(errG))
 
 
         ############################
@@ -219,7 +219,7 @@ for pass_id in range(100):
                     plt.xticks([])
                     plt.yticks([])
                     plt.subplots_adjust(wspace=0.1, hspace=0.1)
-                msg = 'Epoch ID={0} Batch ID={1} \n\n D-Loss={2} G-Loss={3}'.format(pass_id, batch_id, errD.numpy()[0], errG.numpy()[0])
+                msg = 'Epoch ID={0} Batch ID={1} \n\n D-Loss={2} G-Loss={3}'.format(pass_id, batch_id, float(errD), float(errG))
                 print(msg)
                 plt.suptitle(msg,fontsize=20)
                 plt.draw()

ppgan/apps/styleganv2fitting_predictor.py

@@ -144,8 +144,8 @@ class StyleGANv2FittingPredictor(StyleGANv2Predictor):
             optimizer.step()
 
             pbar.set_description(
-                (f"perceptual: {p_loss.numpy()[0]:.4f}; "
-                 f"mse: {mse_loss.numpy()[0]:.4f}; lr: {lr:.4f}"))
+                (f"perceptual: {float(p_loss):.4f}; "
+                 f"mse: {float(mse_loss):.4f}; lr: {lr:.4f}"))
 
         img_gen, _ = generator([latent_n],
                                input_is_latent=True,