Commit @127379ef99d8fdaa089a70f7bec82ae6c4eaf35b

윤영준 2023-06-21

workaround for nn.Sequential can not have multiple input, and I need LSTM for this project

@127379ef99d8fdaa089a70f7bec82ae6c4eaf35b

66fa69f

127379e

model/generator.py

--- model/generator.py

+++ model/generator.py


 from torch import nn
 from torch.nn import functional as F
 
+# nn.Sequential does not handel multiple input by design
+# https://github.com/pytorch/pytorch/issues/19808#
+class mySequential(nn.Sequential):
+    def forward(self, *input):
+        for module in self._modules.values():
+            input = module(*input)
+        return input
 
 def conv3x3(in_ch, out_ch, stride=1, padding=1, groups=1, dilation=1):
     return nn.Conv2d(in_ch, out_ch, kernel_size=3, stride=stride, padding=padding, groups=groups, dilation=dilation)

         self.layers = layers
 
     def forward(self, x):
+        x = self.conv1(x)
         shortcut = x
-        x = self.conv1(shortcut)
         for i, hidden_layer in enumerate(self.conv_hidden):
             x = hidden_layer(x)
             if (i % self.layers == 0) & (i != 0):
-                x = self.relu(x)
-                x += shortcut
+                x = F.relu(x)
+                x = x + shortcut
         return x
 
 

                 dilation=self.dilation
             )
         )
-        self.LSTM = nn.Sequential(
+        self.LSTM = mySequential(
             ConvLSTM(
                 ch=out_ch, kernel_size=kernel_size,
             )

         self.groups = groups
         self.dilation = dilation
         self.repetition = repetition
-        self.generator_block = nn.Sequential(
+        self.generator_block = mySequential(
             GeneratorBlock(blocks=blocks,
                            layers=layers,
                            input_ch=input_ch,

 if __name__ == "__main__":
     from torchinfo import summary
     generator = Generator(3)
-    batch_size = 10
-    summary(generator, input_size=(batch_size, 3, 1920,1080))
+    batch_size = 1
+    summary(generator, input_size=(batch_size, 3, 960,540))

Add a comment

Open 0
Closed 0

List

...	...	@@ -2,6 +2,13 @@
2	2	from torch import nn
3	3	from torch.nn import functional as F
4	4
	5	+# nn.Sequential does not handel multiple input by design
	6	+# https://github.com/pytorch/pytorch/issues/19808#
	7	+class mySequential(nn.Sequential):
	8	+ def forward(self, *input):
	9	+ for module in self._modules.values():
	10	+ input = module(*input)
	11	+ return input
5	12
6	13	def conv3x3(in_ch, out_ch, stride=1, padding=1, groups=1, dilation=1):
7	14	return nn.Conv2d(in_ch, out_ch, kernel_size=3, stride=stride, padding=padding, groups=groups, dilation=dilation)
...	...	@@ -50,13 +57,13 @@
50	57	self.layers = layers
51	58
52	59	def forward(self, x):
	60	+ x = self.conv1(x)
53	61	shortcut = x
54		- x = self.conv1(shortcut)
55	62	for i, hidden_layer in enumerate(self.conv_hidden):
56	63	x = hidden_layer(x)
57	64	if (i % self.layers == 0) & (i != 0):
58		- x = self.relu(x)
59		- x += shortcut
	65	+ x = F.relu(x)
	66	+ x = x + shortcut
60	67	return x
61	68
62	69
...	...	@@ -136,7 +143,7 @@
136	143	dilation=self.dilation
137	144	)
138	145	)
139		- self.LSTM = nn.Sequential(
	146	+ self.LSTM = mySequential(
140	147	ConvLSTM(
141	148	ch=out_ch, kernel_size=kernel_size,
142	149	)
...	...	@@ -168,7 +175,7 @@
168	175	self.groups = groups
169	176	self.dilation = dilation
170	177	self.repetition = repetition
171		- self.generator_block = nn.Sequential(
	178	+ self.generator_block = mySequential(
172	179	GeneratorBlock(blocks=blocks,
173	180	layers=layers,
174	181	input_ch=input_ch,
...	...	@@ -252,5 +259,5 @@
252	259	if __name__ == "__main__":
253	260	from torchinfo import summary
254	261	generator = Generator(3)
255		- batch_size = 10
256		- summary(generator, input_size=(batch_size, 3, 1920,1080))
	262	+ batch_size = 1
	263	+ summary(generator, input_size=(batch_size, 3, 960,540))

Delete comment