NN tutorial - picture and code are confusing at first glance about tutorials

[jmarintur]

Dear @intel-hm, rather than populating the init, perhaps another simple way to connect picture and code could consist in something like this:

import torch
import torch.nn as nn
import torch.nn.functional as F


class Net(nn.Module):

    def __init__(self):
        super(Net, self).__init__()
        # 1 input image channel, 6 output channels, 5x5 square convolution
        # kernel
        self.conv1 = nn.Conv2d(1, 6, 5)
        self.conv2 = nn.Conv2d(6, 16, 5)
        # an affine operation: y = Wx + b
        self.fc1 = nn.Linear(16 * 5 * 5, 120)  # 5*5 from image dimension
        self.fc2 = nn.Linear(120, 84)
        self.fc3 = nn.Linear(84, 10)

    def forward(self, input):
        c1 = F.relu(self.conv1(input))
        # Max pooling over a (2, 2) window
        s2 = F.max_pool2d(c1, (2, 2))
        c3 = F.relu(self.conv2(s2))
        # If the size is a square, you can specify with a single number
        s4 = F.max_pool2d(c3, 2)
        # flatten all s4 dimensions except the batch dimension, before applying fc1
        s4 = torch.flatten(s4, 1)
        f5 = F.relu(self.fc1(s4)) 
        f6 = F.relu(self.fc2(f5))
        output = self.fc3(f6)
        return output

net = Net()
net

Wdyt?

[intel-hm]

I just think that each layer mentioned in the picture should have an corresponding comment in the code. Some of the layers are represented by variables, some are not. Things are further complicated by different names in picture and the private variables of class Net. For someone just starting out this is extremely confusing. Best regards Michael From: Javier Marin Tur ***@***.***> Sent: Saturday, March 30, 2024 10:23 AM To: pytorch/tutorials ***@***.***> Cc: Hebenstreit, Michael ***@***.***>; Mention ***@***.***> Subject: Re: [pytorch/tutorials] NN tutorial - picture and code are confusing at first glance (Issue #2788) Dear @intel-hm<https://github.com/intel-hm>, rather than populating the init, perhaps another simple way to connect picture and code could consist in something like this: import torch import torch.nn as nn import torch.nn.functional as F class Net(nn.Module): def __init__(self): super(Net, self).__init__() # 1 input image channel, 6 output channels, 5x5 square convolution # kernel self.conv1 = nn.Conv2d(1, 6, 5) self.conv2 = nn.Conv2d(6, 16, 5) # an affine operation: y = Wx + b self.fc1 = nn.Linear(16 * 5 * 5, 120) # 5*5 from image dimension self.fc2 = nn.Linear(120, 84) self.fc3 = nn.Linear(84, 10) def forward(self, input): c1 = F.relu(self.conv1(input)) # Max pooling over a (2, 2) window s2 = F.max_pool2d(c1, (2, 2)) c3 = F.relu(self.conv2(s2)) # If the size is a square, you can specify with a single number s4 = F.max_pool2d(c3, 2) # flatten all s4 dimensions except the batch dimension, before applying fc1 s4 = torch.flatten(s4, 1) f5 = F.relu(self.fc1(s4)) f6 = F.relu(self.fc2(f5)) output = self.fc3(f6) return output net = Net() net Wdyt? — Reply to this email directly, view it on GitHub<#2788 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/ASRFKMN4OZINHJHT4PQRVATY23KE7AVCNFSM6AAAAABEGCMSRCVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDAMRYGIYTOMBVGU>. You are receiving this because you were mentioned.Message ID: ***@***.***>

[jmarintur]

We could include the comments you added in your suggestion to the renaming I did within the forward.

[intel-hm]

Please do! Thanks Michael From: Javier Marin Tur ***@***.***> Sent: Saturday, March 30, 2024 12:10 PM To: pytorch/tutorials ***@***.***> Cc: Hebenstreit, Michael ***@***.***>; Mention ***@***.***> Subject: Re: [pytorch/tutorials] NN tutorial - picture and code are confusing at first glance (Issue #2788) We could include the comments you added in your suggestion to the renaming I did within the forward. — Reply to this email directly, view it on GitHub<#2788 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/ASRFKMK2RMHJCGLEZLVFWM3Y23WWZAVCNFSM6AAAAABEGCMSRCVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDAMRYGQYDKOBQGU>. You are receiving this because you were mentioned.Message ID: ***@***.***>

[jmarintur]

Hi @intel-hm, I've just created a PR, I added you as a co-author. I slightly modify the original one.

[intel-hm]

Thank you! From: Javier Marin Tur ***@***.***> Sent: Monday, April 1, 2024 4:41 PM To: pytorch/tutorials ***@***.***> Cc: Hebenstreit, Michael ***@***.***>; Mention ***@***.***> Subject: Re: [pytorch/tutorials] NN tutorial - picture and code are confusing at first glance (Issue #2788) Hi @intel-hm<https://github.com/intel-hm>, I've just created a PR, I added you as a co-author. I slightly modify the original one. — Reply to this email directly, view it on GitHub<#2788 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/ASRFKMKR476GQTASOXSOTMDY3HO7VAVCNFSM6AAAAABEGCMSRCVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDAMZQGY4TSMRZGA>. You are receiving this because you were mentioned.Message ID: ***@***.***>