francescosaveriozuppichini / pytorch-how-and-when-to-use-module-sequential-modulelist-and-moduledict

Code for my medium article

Pytorch: how and when to use Module, Sequential, ModuleList and ModuleDict

Effective way to share, reuse and break down the complexity of your models

Updated at Pytorch 1.5

You can find the code here

Pytorch is an open source deep learning frameworks that provide a smart way to create ML models. Even if the documentation is well made, I still see that most people don't write well and organized code in PyTorch.

Today, we are going to see how to use the three main building blocks of PyTorch: Module, Sequential and ModuleList. We are going to start with an example and iteratively we will make it better.

All these four classes are contained into torch.nn

import torch.nn as nn

# nn.Module
# nn.Sequential
# nn.Module

Module: the main building block

The Module is the main building block, it defines the base class for all neural network and you MUST subclass it.

Let's create a classic CNN classifier as example:

import torch.nn.functional as F

class MyCNNClassifier(nn.Module):
    def __init__(self, in_c, n_classes):
        super().__init__()
        self.conv1 = nn.Conv2d(in_c, 32, kernel_size=3, stride=1, padding=1)
        self.bn1 = nn.BatchNorm2d(32)
        
        self.conv2 = nn.Conv2d(32, 64, kernel_size=3, stride=1, padding=1)
        self.bn2 = nn.BatchNorm2d(64)

        self.fc1 = nn.Linear(64 * 28 * 28, 1024)
        self.fc2 = nn.Linear(1024, n_classes)
        
    def forward(self, x):
        x = self.conv1(x)
        x = self.bn1(x)
        x = F.relu(x)
        
        x = self.conv2(x)
        x = self.bn2(x)
        x = F.relu(x)

        x = x.view(x.size(0), -1) # flat
        
        x = self.fc1(x)
        x = F.sigmoid(x)
        x = self.fc2(x)
        
        return x

model = MyCNNClassifier(1, 10)
print(model)

MyCNNClassifier(
  (conv1): Conv2d(1, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
  (bn1): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
  (conv2): Conv2d(32, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
  (bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
  (fc1): Linear(in_features=50176, out_features=1024, bias=True)
  (fc2): Linear(in_features=1024, out_features=10, bias=True)
)

This is a very simple classifier with an encoding part that uses two layers with 3x3 convs + batchnorm + relu and a decoding part with two linear layers. If you are not new to PyTorch you may have seen this type of coding before, but there are two problems.

If we want to add a layer we have to again write lots of code in the __init__ and in the forward function. Also, if we have some common block that we want to use in another model, e.g. the 3x3 conv + batchnorm + relu, we have to write it again.

Sequential: stack and merge layers

Sequential is a container of Modules that can be stacked together and run at the same time.

You can notice that we have to store into self everything. We can use Sequential to improve our code.

class MyCNNClassifier(nn.Module):
    def __init__(self, in_c, n_classes):
        super().__init__()
        self.conv_block1 = nn.Sequential(
            nn.Conv2d(in_c, 32, kernel_size=3, stride=1, padding=1),
            nn.BatchNorm2d(32),
            nn.ReLU()
        )
        
        self.conv_block2 = nn.Sequential(
            nn.Conv2d(32, 64, kernel_size=3, stride=1, padding=1),
            nn.BatchNorm2d(64),
            nn.ReLU()
        )
        
        self.decoder = nn.Sequential(
            nn.Linear(64 * 28 * 28, 1024),
            nn.Sigmoid(),
            nn.Linear(1024, n_classes)
        )

        
    def forward(self, x):
        x = self.conv_block1(x)
        x = self.conv_block2(x)

        x = x.view(x.size(0), -1) # flat
        
        x = self.decoder(x)
        
        return x

model = MyCNNClassifier(1, 10)
print(model)

MyCNNClassifier(
  (conv_block1): Sequential(
    (0): Conv2d(1, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (1): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (2): ReLU()
  )
  (conv_block2): Sequential(
    (0): Conv2d(32, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (2): ReLU()
  )
  (decoder): Sequential(
    (0): Linear(in_features=50176, out_features=1024, bias=True)
    (1): Sigmoid()
    (2): Linear(in_features=1024, out_features=10, bias=True)
  )
)

Much Better uhu?

Did you notice that conv_block1 and conv_block2 looks almost the same? We could create a function that reteurns a nn.Sequential to even simplify the code!

def conv_block(in_f, out_f, *args, **kwargs):
    return nn.Sequential(
        nn.Conv2d(in_f, out_f, *args, **kwargs),
        nn.BatchNorm2d(out_f),
        nn.ReLU()
    )

Then we can just call this function in our Module

class MyCNNClassifier(nn.Module):
    def __init__(self, in_c, n_classes):
        super().__init__()
        self.conv_block1 = conv_block(in_c, 32, kernel_size=3, padding=1)
        
        self.conv_block2 = conv_block(32, 64, kernel_size=3, padding=1)

        
        self.decoder = nn.Sequential(
            nn.Linear(64 * 28 * 28, 1024),
            nn.Sigmoid(),
            nn.Linear(1024, n_classes)
        )

        
    def forward(self, x):
        x = self.conv_block1(x)
        x = self.conv_block2(x)

        x = x.view(x.size(0), -1) # flat
        
        x = self.decoder(x)
        
        return x

model = MyCNNClassifier(1, 10)
print(model)

MyCNNClassifier(
  (conv_block1): Sequential(
    (0): Conv2d(1, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (1): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (2): ReLU()
  )
  (conv_block2): Sequential(
    (0): Conv2d(32, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (2): ReLU()
  )
  (decoder): Sequential(
    (0): Linear(in_features=50176, out_features=1024, bias=True)
    (1): Sigmoid()
    (2): Linear(in_features=1024, out_features=10, bias=True)
  )
)

Even cleaner! Still conv_block1 and conv_block2 are almost the same! We can merge them using nn.Sequential

class MyCNNClassifier(nn.Module):
    def __init__(self, in_c, n_classes):
        super().__init__()
        self.encoder = nn.Sequential(
            conv_block(in_c, 32, kernel_size=3, padding=1),
            conv_block(32, 64, kernel_size=3, padding=1)
        )

        
        self.decoder = nn.Sequential(
            nn.Linear(64 * 28 * 28, 1024),
            nn.Sigmoid(),
            nn.Linear(1024, n_classes)
        )

        
    def forward(self, x):
        x = self.encoder(x)
        
        x = x.view(x.size(0), -1) # flat
        
        x = self.decoder(x)
        
        return x

model = MyCNNClassifier(1, 10)
print(model)

MyCNNClassifier(
  (encoder): Sequential(
    (0): Sequential(
      (0): Conv2d(1, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (1): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (2): ReLU()
    )
    (1): Sequential(
      (0): Conv2d(32, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (2): ReLU()
    )
  )
  (decoder): Sequential(
    (0): Linear(in_features=50176, out_features=1024, bias=True)
    (1): Sigmoid()
    (2): Linear(in_features=1024, out_features=10, bias=True)
  )
)

self.encoder now holds booth conv_block. We have decoupled logic for our model and make it easier to read and reuse. Our conv_block function can be imported and used in another model.

Dynamic Sequential: create multiple layers at once

What if we can to add a new layers in self.encoder, hardcoded them is not convinient:

self.encoder = nn.Sequential(
            conv_block(in_c, 32, kernel_size=3, padding=1),
            conv_block(32, 64, kernel_size=3, padding=1),
            conv_block(64, 128, kernel_size=3, padding=1),
            conv_block(128, 256, kernel_size=3, padding=1),

        )

Would it be nice if we can define the sizes as an array and automatically create all the layers without writing each one of them? Fortunately we can create an array and pass it to Sequential

class MyCNNClassifier(nn.Module):
    def __init__(self, in_c, n_classes):
        super().__init__()
        self.enc_sizes = [in_c, 32, 64]
        
        conv_blocks = [conv_block(in_f, out_f, kernel_size=3, padding=1) 
                       for in_f, out_f in zip(self.enc_sizes, self.enc_sizes[1:])]
        
        self.encoder = nn.Sequential(*conv_blocks)

        
        self.decoder = nn.Sequential(
            nn.Linear(64 * 28 * 28, 1024),
            nn.Sigmoid(),
            nn.Linear(1024, n_classes)
        )

        
    def forward(self, x):
        x = self.encoder(x)
        
        x = x.view(x.size(0), -1) # flat
        
        x = self.decoder(x)
        
        return x

model = MyCNNClassifier(1, 10)
print(model)

MyCNNClassifier(
  (encoder): Sequential(
    (0): Sequential(
      (0): Conv2d(1, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (1): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (2): ReLU()
    )
    (1): Sequential(
      (0): Conv2d(32, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (2): ReLU()
    )
  )
  (decoder): Sequential(
    (0): Linear(in_features=50176, out_features=1024, bias=True)
    (1): Sigmoid()
    (2): Linear(in_features=1024, out_features=10, bias=True)
  )
)

Let's break it down. We created an array self.enc_sizes that holds the sizes of our encoder. Then we create an array conv_blocks by iterating the sizes. Since we have to give booth a in size and an outsize for each layer we ziped the size'array with itself by shifting it by one.

Just to be clear, take a look at the following example:

sizes = [1, 32, 64]

for in_f,out_f in zip(sizes, sizes[1:]):
    print(in_f,out_f)

1 32
32 64

Then, since Sequential does not accept a list, we decompose it by using the * operator.

Tada! Now if we just want to add a size, we can easily add a new number to the list. It is a common practice to make the size a parameter.

class MyCNNClassifier(nn.Module):
    def __init__(self, in_c, enc_sizes, n_classes):
        super().__init__()
        self.enc_sizes = [in_c, *enc_sizes]
        
        conv_blocks = [conv_block(in_f, out_f, kernel_size=3, padding=1) 
                       for in_f, out_f in zip(self.enc_sizes, self.enc_sizes[1:])]
        
        self.encoder = nn.Sequential(*conv_blocks)

        
        self.decoder = nn.Sequential(
            nn.Linear(64 * 28 * 28, 1024),
            nn.Sigmoid(),
            nn.Linear(1024, n_classes)
        )

        
    def forward(self, x):
        x = self.encoder(x)
        
        x = x.view(x.size(0), -1) # flat
        
        x = self.decoder(x)
        
        return x

model = MyCNNClassifier(1, [32,64, 128], 10)
print(model)

MyCNNClassifier(
  (encoder): Sequential(
    (0): Sequential(
      (0): Conv2d(1, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (1): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (2): ReLU()
    )
    (1): Sequential(
      (0): Conv2d(32, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (2): ReLU()
    )
    (2): Sequential(
      (0): Conv2d(64, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (2): ReLU()
    )
  )
  (decoder): Sequential(
    (0): Linear(in_features=50176, out_features=1024, bias=True)
    (1): Sigmoid()
    (2): Linear(in_features=1024, out_features=10, bias=True)
  )
)

We can do the same for the decoder part

def dec_block(in_f, out_f):
    return nn.Sequential(
        nn.Linear(in_f, out_f),
        nn.Sigmoid()
    )

class MyCNNClassifier(nn.Module):
    def __init__(self, in_c, enc_sizes, dec_sizes,  n_classes):
        super().__init__()
        self.enc_sizes = [in_c, *enc_sizes]
        self.dec_sizes = [64 * 28 * 28, *dec_sizes]

        conv_blocks = [conv_block(in_f, out_f, kernel_size=3, padding=1) 
                       for in_f, out_f in zip(self.enc_sizes, self.enc_sizes[1:])]
        
        self.encoder = nn.Sequential(*conv_blocks)

        
        dec_blocks = [dec_block(in_f, out_f) 
                       for in_f, out_f in zip(self.dec_sizes, self.dec_sizes[1:])]
        
        self.decoder = nn.Sequential(*dec_blocks)
        
        self.last = nn.Linear(self.dec_sizes[-1], n_classes)

        
    def forward(self, x):
        x = self.encoder(x)
        
        x = x.view(x.size(0), -1) # flat
        
        x = self.decoder(x)
        
        return x

model = MyCNNClassifier(1, [32,64], [1024, 512], 10)
print(model)

MyCNNClassifier(
  (encoder): Sequential(
    (0): Sequential(
      (0): Conv2d(1, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (1): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (2): ReLU()
    )
    (1): Sequential(
      (0): Conv2d(32, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (2): ReLU()
    )
  )
  (decoder): Sequential(
    (0): Sequential(
      (0): Linear(in_features=50176, out_features=1024, bias=True)
      (1): Sigmoid()
    )
    (1): Sequential(
      (0): Linear(in_features=1024, out_features=512, bias=True)
      (1): Sigmoid()
    )
  )
  (last): Linear(in_features=512, out_features=10, bias=True)
)

We followed the same pattern, we create a new block for the decoding part, linear + sigmoid, and we pass an array with the sizes. We had to add a self.last since we do not want to activate the output

Now, we can even break down our model in two! Encoder + Decoder

class MyEncoder(nn.Module):
    def __init__(self, enc_sizes):
        super().__init__()
        self.conv_blocks = nn.Sequential(*[conv_block(in_f, out_f, kernel_size=3, padding=1) 
                       for in_f, out_f in zip(enc_sizes, enc_sizes[1:])])

        def forward(self, x):
            return self.conv_blocks(x)
        
class MyDecoder(nn.Module):
    def __init__(self, dec_sizes, n_classes):
        super().__init__()
        self.dec_blocks = nn.Sequential(*[dec_block(in_f, out_f) 
                       for in_f, out_f in zip(dec_sizes, dec_sizes[1:])])
        self.last = nn.Linear(dec_sizes[-1], n_classes)

    def forward(self, x):
        return self.dec_blocks()
    
    
class MyCNNClassifier(nn.Module):
    def __init__(self, in_c, enc_sizes, dec_sizes,  n_classes):
        super().__init__()
        self.enc_sizes = [in_c, *enc_sizes]
        self.dec_sizes = [self.enc_sizes[-1] * 28 * 28, *dec_sizes]

        self.encoder = MyEncoder(self.enc_sizes)
        
        self.decoder = MyDecoder(self.dec_sizes, n_classes)
        
    def forward(self, x):
        x = self.encoder(x)
        
        x = x.flatten(1) # flat
        
        x = self.decoder(x)
        
        return x

model = MyCNNClassifier(1, [32,64], [1024, 512], 10)
print(model)

MyCNNClassifier(
  (encoder): MyEncoder(
    (conv_blocks): Sequential(
      (0): Sequential(
        (0): Conv2d(1, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
        (1): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (2): ReLU()
      )
      (1): Sequential(
        (0): Conv2d(32, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
        (1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (2): ReLU()
      )
    )
  )
  (decoder): MyDecoder(
    (dec_blocks): Sequential(
      (0): Sequential(
        (0): Linear(in_features=50176, out_features=1024, bias=True)
        (1): Sigmoid()
      )
      (1): Sequential(
        (0): Linear(in_features=1024, out_features=512, bias=True)
        (1): Sigmoid()
      )
    )
    (last): Linear(in_features=512, out_features=10, bias=True)
  )
)

Be aware that MyEncoder and MyDecoder could also be functions that returns a nn.Sequential. I prefer to use the first pattern for models and the second for building blocks.

By diving our module into submodules it is easier to share the code, debug it and test it.

ModuleList : when we need to iterate

ModuleList allows you to store Module as a list. It can be useful when you need to iterate through layer and store/use some information, like in U-net.

The main difference between Sequential is that ModuleList have not a forward method so the inner layers are not connected. Assuming we need each output of each layer in the decoder, we can store it by:

class MyModule(nn.Module):
    def __init__(self, sizes):
        super().__init__()
        self.layers = nn.ModuleList([nn.Linear(in_f, out_f) for in_f, out_f in zip(sizes, sizes[1:])])
        self.trace = []
        
    def forward(self,x):
        for layer in self.layers:
            x = layer(x)
            self.trace.append(x)
        return x

model = MyModule([1, 16, 32])
import torch

model(torch.rand((4,1)))

[print(trace.shape) for trace in model.trace]

torch.Size([4, 16])
torch.Size([4, 32])





[None, None]

ModuleDict: when we need to choose

What if we want to switch to LearkyRelu in our conv_block? We can use ModuleDict to create a dictionary of Module and dynamically switch Module when we want

def conv_block(in_f, out_f, activation='relu', *args, **kwargs):
    
    activations = nn.ModuleDict([
                ['lrelu', nn.LeakyReLU()],
                ['relu', nn.ReLU()]
    ])
    
    return nn.Sequential(
        nn.Conv2d(in_f, out_f, *args, **kwargs),
        nn.BatchNorm2d(out_f),
        activations[activation]
    )

print(conv_block(1, 32,'lrelu', kernel_size=3, padding=1))
print(conv_block(1, 32,'relu', kernel_size=3, padding=1))

Sequential(
  (0): Conv2d(1, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
  (1): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
  (2): LeakyReLU(negative_slope=0.01)
)
Sequential(
  (0): Conv2d(1, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
  (1): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
  (2): ReLU()
)

Final implementation

Let's wrap it up everything!

def conv_block(in_f, out_f, activation='relu', *args, **kwargs):
    activations = nn.ModuleDict([
                ['lrelu', nn.LeakyReLU()],
                ['relu', nn.ReLU()]
    ])
    
    return nn.Sequential(
        nn.Conv2d(in_f, out_f, *args, **kwargs),
        nn.BatchNorm2d(out_f),
        activations[activation]
    )

def dec_block(in_f, out_f):
    return nn.Sequential(
        nn.Linear(in_f, out_f),
        nn.Sigmoid()
    )

class MyEncoder(nn.Module):
    def __init__(self, enc_sizes, *args, **kwargs):
        super().__init__()
        self.conv_blocks = nn.Sequential(*[conv_block(in_f, out_f, kernel_size=3, padding=1, *args, **kwargs) 
                       for in_f, out_f in zip(enc_sizes, enc_sizes[1:])])
        
        def forward(self, x):
            return self.conv_blocks(x)
        
class MyDecoder(nn.Module):
    def __init__(self, dec_sizes, n_classes):
        super().__init__()
        self.dec_blocks = nn.Sequential(*[dec_block(in_f, out_f) 
                       for in_f, out_f in zip(dec_sizes, dec_sizes[1:])])
        self.last = nn.Linear(dec_sizes[-1], n_classes)

    def forward(self, x):
        return self.dec_blocks()
    
    
class MyCNNClassifier(nn.Module):
    def __init__(self, in_c, enc_sizes, dec_sizes,  n_classes, activation='relu'):
        super().__init__()
        self.enc_sizes = [in_c, *enc_sizes]
        self.dec_sizes = [32 * 28 * 28, *dec_sizes]

        self.encoder = MyEncoder(self.enc_sizes, activation=activation)
        
        self.decoder = MyDecoder(dec_sizes, n_classes)
        
    def forward(self, x):
        x = self.encoder(x)
        
        x = x.flatten(1) # flat
        
        x = self.decoder(x)
        
        return x

model = MyCNNClassifier(1, [32,64], [1024, 512], 10, activation='lrelu')
print(model)

MyCNNClassifier(
  (encoder): MyEncoder(
    (conv_blocks): Sequential(
      (0): Sequential(
        (0): Conv2d(1, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
        (1): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (2): LeakyReLU(negative_slope=0.01)
      )
      (1): Sequential(
        (0): Conv2d(32, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
        (1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (2): LeakyReLU(negative_slope=0.01)
      )
    )
  )
  (decoder): MyDecoder(
    (dec_blocks): Sequential(
      (0): Sequential(
        (0): Linear(in_features=1024, out_features=512, bias=True)
        (1): Sigmoid()
      )
    )
    (last): Linear(in_features=512, out_features=10, bias=True)
  )
)

Conclusion

So, in summary.

• Use Module when you have a big block compose of multiple smaller blocks
• Use Sequential when you want to create a small block from layers
• Use ModuleList when you need to iterate through some layers or building blocks and do something
• Use ModuleDict when you need to parametise some blocks of your model, for example an activation function

That's all folks!

Thank you for reading

pytorch-how-and-when-to-use-module-sequential-modulelist-and-moduledict's Issues

Maybe worth automate one entry to have it really fully automated...

This is just a suggestion. There is one entry which would depend on the data itself. Maybe it is worth calculating it automatically as well? Any ideas on how to do it?

class MyCNNClassifier(nn.Module):
    def __init__(self, in_c, enc_sizes, dec_sizes,  n_classes, activation='relu'):
        super().__init__()
        self.enc_sizes = [in_c, *enc_sizes]
        self.dec_sizes = [32 * 28 * 28, *dec_sizes] # here: 32 * 28 *28

        self.encoder = MyEncoder(self.enc_sizes, activation=activation)
        
        self.decoder = MyDecoder(dec_sizes, n_classes)
        
    def forward(self, x):
        x = self.encoder(x)
        
        x = x.flatten(1) # flat
        
        x = self.decoder(x)
        
        return x

Insignificant typo in Conclusions

There is the word parametise which should be parameterize.

Is the number wrong?

Maybe I'm wrong, but I think the number is 64:

class MyCNNClassifier(nn.Module):
    def __init__(self, in_c, n_classes):
        super().__init__()
        self.conv1 = nn.Conv2d(in_c, 32, kernel_size=3, stride=1, padding=1)
        self.bn1 = nn.BatchNorm2d(32)
        
        self.conv2 = nn.Conv2d(32, 64, kernel_size=3, stride=1, padding=1)
        self.bn2 = nn.BatchNorm2d(64) # HERE

        self.fc1 = nn.Linear(64 * 28 * 28, 1024) # HERE
        self.fc2 = nn.Linear(1024, n_classes)

conv_blokcs aesthetically doesn't look good :)

It's not an issue, but from aesthetical point of view of a perfectionist, it looks better with conv_blocks instead of conv_blokcs :)

class MyCNNClassifier(nn.Module):
    def __init__(self, in_c, enc_sizes, n_classes):
        super().__init__()
        self.enc_sizes = [in_c, *enc_sizes]
        
        conv_blokcs = [conv_block(in_f, out_f, kernel_size=3, padding=1)  # here
                       for in_f, out_f in zip(self.enc_sizes, self.enc_sizes[1:])]
        
        self.encoder = nn.Sequential(*conv_blokcs) # here

        
        self.decoder = nn.Sequential(
            nn.Linear(32 * 28 * 28, 1024),
            nn.Sigmoid(),
            nn.Linear(1024, n_classes)
        )

        
    def forward(self, x):
        x = self.encoder(x)
        
        x = x.view(x.size(0), -1) # flat
        
        x = self.decoder(x)
        
        return x

`conv_blocks` instead of `conv_blokcs`: one chunk of code left to correct this typo

I know you looked through all the chunks of codes and corrected it, but there is one left.
Quote: "It's not an issue, but from aesthetical point of view of a perfectionist, it looks better with conv_blocks instead of conv_blokcs :)"

class MyCNNClassifier(nn.Module):
    def __init__(self, in_c, enc_sizes, n_classes):
        super().__init__()
        self.enc_sizes = [in_c, *enc_sizes]
        
        conv_blokcs = [conv_block(in_f, out_f, kernel_size=3, padding=1)  # here
                       for in_f, out_f in zip(self.enc_sizes, self.enc_sizes[1:])]
        
        self.encoder = nn.Sequential(*conv_blokcs) # here

        
        self.decoder = nn.Sequential(
            nn.Linear(32 * 28 * 28, 1024),
            nn.Sigmoid(),
            nn.Linear(1024, n_classes)
        )

        
    def forward(self, x):
        x = self.encoder(x)
        
        x = x.view(x.size(0), -1) # flat
        
        x = self.decoder(x)
        
        return x

Forgotten line?

I think it is forgotten to put self.last() in forward

def dec_block(in_f, out_f):
    return nn.Sequential(
        nn.Linear(in_f, out_f),
        nn.Sigmoid()
    )

class MyCNNClassifier(nn.Module):
    def __init__(self, in_c, enc_sizes, dec_sizes,  n_classes):
        super().__init__()
        self.enc_sizes = [in_c, *enc_sizes]
        self.dec_sizes = [32 * 28 * 28, *dec_sizes]

        conv_blokcs = [conv_block(in_f, out_f, kernel_size=3, padding=1) 
                       for in_f, out_f in zip(self.enc_sizes, self.enc_sizes[1:])]
        
        self.encoder = nn.Sequential(*conv_blokcs)

        
        dec_blocks = [dec_block(in_f, out_f) 
                       for in_f, out_f in zip(self.dec_sizes, self.dec_sizes[1:])]
        
        self.decoder = nn.Sequential(*dec_blocks)
        
        self.last = nn.Linear(self.dec_sizes[-1], n_classes)

        
    def forward(self, x):
        x = self.encoder(x)
        
        x = x.view(x.size(0), -1) # flat
        
        x = self.decoder(x)

        x = self.last(x) # here
        
        return x

Should it be 'return self.dec_blocks(x)'?

Should it be 'return self.dec_blocks(x)'?
There are two places with this code-line.

class MyDecoder(nn.Module):
    def __init__(self, dec_sizes, n_classes):
        super().__init__()
        self.dec_blocks = nn.Sequential(*[dec_block(in_f, out_f) 
                       for in_f, out_f in zip(dec_sizes, dec_sizes[1:])])
        self.last = nn.Linear(dec_sizes[-1], n_classes)

    def forward(self, x):
        return self.dec_blocks() # Here

Should it be 'self.dec_sizes'?

Should it be 'self.dec_sizes'? (there two chunks of codes to correct)

class MyCNNClassifier(nn.Module):
    def __init__(self, in_c, enc_sizes, dec_sizes,  n_classes):
        super().__init__()
        self.enc_sizes = [in_c, *enc_sizes]
        self.dec_sizes = [32 * 28 * 28, *dec_sizes]

        self.encoder = MyEncoder(self.enc_sizes)
        
        self.decoder = MyDecoder(dec_sizes, n_classes) # here
        
    def forward(self, x):
        x = self.encoder(x)
        
        x = x.flatten(1) # flat
        
        x = self.decoder(x)
        
        return x

Should it be nn.ModuleList?

Should it be nn.ModuleList?
Also, you actually consider ModuleDict as well. Just a suggestion, maybe it is worth mentioning in the beginning not three main building blocks but rather four, just for the sake of consistency. And then add the fourth line below as well.

import torch.nn as nn

# nn.Module
# nn.Sequential
# nn.Module # here
# nn.ModuleDict # here

Update notebook

Update notebook to reflect the fixed in the .md