Update: Optimize some code.

model/network.py

@@ -56,7 +56,7 @@ def __init__(self, in_channel=3, out_channel=3, channel=None, time_channel=256,
         self.outc = nn.Conv2d(in_channels=channel[0], out_channels=out_channel, kernel_size=1)
 
         if num_classes is not None:
-            self.label_emb = nn.Embedding(num_classes, time_channel)
+            self.label_emb = nn.Embedding(num_embeddings=num_classes, embedding_dim=time_channel)
 
     def pos_encoding(self, time, channels):
         """
@@ -65,10 +65,11 @@ def pos_encoding(self, time, channels):
         :param channels: Channels
         :return: pos_enc
         """
-        inv_freq = 1.0 / (10000 ** (torch.arange(0, channels, 2, device=self.device).float() / channels))
-        pos_enc_a = torch.sin(time.repeat(1, channels // 2) * inv_freq)
-        pos_enc_b = torch.cos(time.repeat(1, channels // 2) * inv_freq)
-        pos_enc = torch.cat([pos_enc_a, pos_enc_b], dim=-1)
+        inv_freq = 1.0 / (10000 ** (torch.arange(start=0, end=channels, step=2, device=self.device).float() / channels))
+        inv_freq_value = time.repeat(1, channels // 2) * inv_freq
+        pos_enc_a = torch.sin(input=inv_freq_value)
+        pos_enc_b = torch.cos(input=inv_freq_value)
+        pos_enc = torch.cat(tensors=[pos_enc_a, pos_enc_b], dim=-1)
         return pos_enc
 
     def forward(self, x, time, y=None):