I have been reading most of the questions regarding the nn.ModuleList() and I thought I understood how to use it. But, apparently, I am missing something here.
I am creating a network based on two nn.ModuleList() and use one after another, then i want to see if it is learning anything, so based on the pytorch tutorial I tried it on CIFA10 based on the steps that is provided in Training a Classifier. every step is the same except my netweok, but the network does not learn anything, so i assume the issue is the way that im using the nn.ModuleList().
I am not sure where/what im doing wrong, everything looks okay.
Can someone please tell me why it is not working and what im doing wrong here 
here is network
class Prediction(nn.Module):
def __init__(self):
super(Prediction, self).__init__()
self.base = self.VGG16()
self.prediction = self.Extention()
def forward(self, x):
for i, name in enumerate(self.base):
x = self.base[i](x)
# Copy x to x22
x22 = x.clone()
S2 = x22.view(x22.size(0),512)
S2 = F.relu(self.prediction[0](S2))
S2 = F.relu(self.prediction[1](S2))
S2 = self.prediction[2](S2)
return S2
def VGG16(self):
cfg = [64, 64, 'M', 128, 128, 'M', 256, 256, 256, 'M', 512, 512, 512, 'M', 512, 512, 512, 'M']
layers = nn.ModuleList()
in_channels = 3
for x in cfg:
if x == 'M':
layers.append(nn.MaxPool2d(kernel_size=2, stride=2, ceil_mode=True))
else:
layers.append(nn.Conv2d(in_channels, x, kernel_size=3, padding=1))
layers.append(nn.ReLU(True))
in_channels = x
return layers
def Extention(self):
prediction = nn.ModuleList()
FC1 = nn.Linear(512, 120)
FC2 = nn.Linear(120, 84)
FC3 = nn.Linear(84, 10)
prediction.append(FC1)
prediction.append(FC2)
prediction.append(FC3)
return prediction
net = Prediction()
the self.base is the VGG16 without the classifier layer.
and the self.Extension is the classifier. 
Here is also the whole code:
import torchvision
import torchvision.transforms as transforms
bb = 4
transform = transforms.Compose(
[transforms.ToTensor(),
transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))])
trainset = torchvision.datasets.CIFAR10(root='./data', train=True,
download=True, transform=transform)
trainloader = torch.utils.data.DataLoader(trainset, batch_size=bb,
shuffle=True, num_workers=2)
testset = torchvision.datasets.CIFAR10(root='./data', train=False,
download=True, transform=transform)
testloader = torch.utils.data.DataLoader(testset, batch_size=bb,
shuffle=False, num_workers=2)
classes = ('plane', 'car', 'bird', 'cat',
'deer', 'dog', 'frog', 'horse', 'ship', 'truck')
import matplotlib.pyplot as plt
import numpy as np
# functions to show an image
def imshow(img):
img = img / 2 + 0.5 # unnormalize
npimg = img.numpy()
plt.imshow(np.transpose(npimg, (1, 2, 0)))
plt.show()
# get some random training images
dataiter = iter(trainloader)
images, labels = dataiter.next()
import torch.nn as nn
import torch.nn.functional as F
class Prediction(nn.Module):
def __init__(self):
super(Prediction, self).__init__()
self.base = self.VGG16()
self.prediction = self.Extention()
def forward(self, x):
for i, name in enumerate(self.base):
x = self.base[i](x)
# Copy x to x22
x22 = x.clone()
S2 = x22.view(x22.size(0),512)
S2 = F.relu(self.prediction[0](S2))
S2 = F.relu(self.prediction[1](S2))
S2 = self.prediction[2](S2)
return S2
def VGG16(self):
cfg = [64, 64, 'M', 128, 128, 'M', 256, 256, 256, 'M', 512, 512, 512, 'M', 512, 512, 512, 'M']
layers = nn.ModuleList()
in_channels = 3
for x in cfg:
if x == 'M':
layers.append(nn.MaxPool2d(kernel_size=2, stride=2, ceil_mode=True))
else:
layers.append(nn.Conv2d(in_channels, x, kernel_size=3, padding=1))
layers.append(nn.ReLU(True))
in_channels = x
return layers
def Extention(self):
prediction = nn.ModuleList()
FC1 = nn.Linear(512, 120)
FC2 = nn.Linear(120, 84)
FC3 = nn.Linear(84, 10)
prediction.append(FC1)
prediction.append(FC2)
prediction.append(FC3)
return prediction
net = Prediction()
import torch.optim as optim
criterion = nn.CrossEntropyLoss()
optimizer = optim.SGD(net.parameters(), lr=0.001, momentum=0.9)
for epoch in range(2): # loop over the dataset multiple times
running_loss = 0.0
for i, data in enumerate(trainloader, 0):
# get the inputs
inputs, labels = data
# zero the parameter gradients
optimizer.zero_grad()
# forward + backward + optimize
outputs = net(inputs)
loss = criterion(outputs, labels)
loss.backward()
optimizer.step()
# print statistics
running_loss += loss.item()
if i % 200 == 199: # print every 2000 mini-batches
print('[%d, %5d] loss: %.3f' %
(epoch + 1, i + 1, running_loss / 2000))
running_loss = 0.0
print('Finished Training')
the output is:
Files already downloaded and verified
Files already downloaded and verified
[1, 200] loss: 0.230
[1, 400] loss: 0.230
[1, 600] loss: 0.230
[1, 800] loss: 0.230
[1, 1000] loss: 0.230
[1, 1200] loss: 0.230
[1, 1400] loss: 0.230
[1, 1600] loss: 0.230
[1, 1800] loss: 0.230
[1, 2000] loss: 0.230
[1, 2200] loss: 0.230
[1, 2400] loss: 0.230
[1, 2600] loss: 0.230
[1, 2800] loss: 0.230
[1, 3000] loss: 0.230
[1, 3200] loss: 0.230
[1, 3400] loss: 0.230
[1, 3600] loss: 0.230
[1, 3800] loss: 0.230
[1, 4000] loss: 0.230
[1, 4200] loss: 0.230
[1, 4400] loss: 0.230
[1, 4600] loss: 0.230
so as we can see no loss is decreasing