diff --git a/models.py b/models.py
index da233d02d..c1cf2117d 100644
--- a/models.py
+++ b/models.py
@@ -85,7 +85,7 @@ def __init__(self, h):
         for i, (u, k) in enumerate(zip(h.upsample_rates, h.upsample_kernel_sizes)):
             self.ups.append(weight_norm(
                 ConvTranspose1d(h.upsample_initial_channel//(2**i), h.upsample_initial_channel//(2**(i+1)),
-                                k, u, padding=(k-u)//2)))
+                                k, u, padding=(u//2 + u%2), output_padding=u%2)))
 
         self.resblocks = nn.ModuleList()
         for i in range(len(self.ups)):
diff --git a/train.py b/train.py
index 3b5509428..5c2733a45 100644
--- a/train.py
+++ b/train.py
@@ -27,6 +27,7 @@ def train(rank, a, h):
                            world_size=h.dist_config['world_size'] * h.num_gpus, rank=rank)
 
     torch.cuda.manual_seed(h.seed)
+    torch.cuda.set_device(rank)
     device = torch.device('cuda:{:d}'.format(rank))
 
     generator = Generator(h).to(device)