src.model 

 1import torch
 2from torch import nn
 3
 4
 5class BasicCNN(nn.Module):
 6    """Example Convolutional Neural Network (CNN) for MNIST identification."""
 7
 8    def __init__(self):
 9        super().__init__()
10
11        self.conv1 = nn.Conv2d(in_channels=1, out_channels=10, kernel_size=3, padding="valid")
12        self.conv2 = nn.Conv2d(in_channels=10, out_channels=7, kernel_size=3, padding="valid")
13        self.conv3 = nn.Conv2d(in_channels=7, out_channels=5, kernel_size=3, padding="valid")
14        self.conv4 = nn.Conv2d(in_channels=5, out_channels=1, kernel_size=3, padding="valid")
15
16        self.act1 = nn.ReLU()
17        self.act2 = nn.ReLU()
18        self.act3 = nn.ReLU()
19        self.act4 = nn.ReLU()
20
21        self.flatten = nn.Flatten(start_dim=1, end_dim=-1)
22        self.linear1 = nn.Linear(in_features=400, out_features=10)
23
24    def forward(self, x: torch.Tensor) -> tuple[torch.Tensor, torch.Tensor]:
25        """Returns prediction as well as the embedding (one layer before final prediction).  The embedding is used for visualizing the projections.
26
27        Args:
28            x (torch.Tensor): The input image.
29
30        Returns:
31            tuple[torch.Tensor, torch.Tensor]: Prediction and embedding tensors.
32        """
33        x = self.conv1(x)
34        x = self.act1(x)
35        x = self.conv2(x)
36        x = self.act2(x)
37        x = self.conv3(x)
38        x = self.act3(x)
39        x = self.conv4(x)
40        x = self.act4(x)
41        e = self.flatten(x)
42        x = self.linear1(e)
43
44        return x, e
class BasicCNN(torch.nn.modules.module.Module): 
 6class BasicCNN(nn.Module):
 7    """Example Convolutional Neural Network (CNN) for MNIST identification."""
 8
 9    def __init__(self):
10        super().__init__()
11
12        self.conv1 = nn.Conv2d(in_channels=1, out_channels=10, kernel_size=3, padding="valid")
13        self.conv2 = nn.Conv2d(in_channels=10, out_channels=7, kernel_size=3, padding="valid")
14        self.conv3 = nn.Conv2d(in_channels=7, out_channels=5, kernel_size=3, padding="valid")
15        self.conv4 = nn.Conv2d(in_channels=5, out_channels=1, kernel_size=3, padding="valid")
16
17        self.act1 = nn.ReLU()
18        self.act2 = nn.ReLU()
19        self.act3 = nn.ReLU()
20        self.act4 = nn.ReLU()
21
22        self.flatten = nn.Flatten(start_dim=1, end_dim=-1)
23        self.linear1 = nn.Linear(in_features=400, out_features=10)
24
25    def forward(self, x: torch.Tensor) -> tuple[torch.Tensor, torch.Tensor]:
26        """Returns prediction as well as the embedding (one layer before final prediction).  The embedding is used for visualizing the projections.
27
28        Args:
29            x (torch.Tensor): The input image.
30
31        Returns:
32            tuple[torch.Tensor, torch.Tensor]: Prediction and embedding tensors.
33        """
34        x = self.conv1(x)
35        x = self.act1(x)
36        x = self.conv2(x)
37        x = self.act2(x)
38        x = self.conv3(x)
39        x = self.act3(x)
40        x = self.conv4(x)
41        x = self.act4(x)
42        e = self.flatten(x)
43        x = self.linear1(e)
44
45        return x, e

Example Convolutional Neural Network (CNN) for MNIST identification.

BasicCNN() 
 9    def __init__(self):
10        super().__init__()
11
12        self.conv1 = nn.Conv2d(in_channels=1, out_channels=10, kernel_size=3, padding="valid")
13        self.conv2 = nn.Conv2d(in_channels=10, out_channels=7, kernel_size=3, padding="valid")
14        self.conv3 = nn.Conv2d(in_channels=7, out_channels=5, kernel_size=3, padding="valid")
15        self.conv4 = nn.Conv2d(in_channels=5, out_channels=1, kernel_size=3, padding="valid")
16
17        self.act1 = nn.ReLU()
18        self.act2 = nn.ReLU()
19        self.act3 = nn.ReLU()
20        self.act4 = nn.ReLU()
21
22        self.flatten = nn.Flatten(start_dim=1, end_dim=-1)
23        self.linear1 = nn.Linear(in_features=400, out_features=10)

Initializes internal Module state, shared by both nn.Module and ScriptModule.

conv1
conv2
conv3
conv4
act1
act2
act3
act4
flatten
linear1
def forward(self, x: torch.Tensor) -> tuple[torch.Tensor, torch.Tensor]: 
25    def forward(self, x: torch.Tensor) -> tuple[torch.Tensor, torch.Tensor]:
26        """Returns prediction as well as the embedding (one layer before final prediction).  The embedding is used for visualizing the projections.
27
28        Args:
29            x (torch.Tensor): The input image.
30
31        Returns:
32            tuple[torch.Tensor, torch.Tensor]: Prediction and embedding tensors.
33        """
34        x = self.conv1(x)
35        x = self.act1(x)
36        x = self.conv2(x)
37        x = self.act2(x)
38        x = self.conv3(x)
39        x = self.act3(x)
40        x = self.conv4(x)
41        x = self.act4(x)
42        e = self.flatten(x)
43        x = self.linear1(e)
44
45        return x, e

Returns prediction as well as the embedding (one layer before final prediction). The embedding is used for visualizing the projections.

Arguments:
  • x (torch.Tensor): The input image.
Returns:

tuple[torch.Tensor, torch.Tensor]: Prediction and embedding tensors.

Inherited Members
torch.nn.modules.module.Module
dump_patches
training
call_super_init
register_buffer
register_parameter
add_module
register_module
get_submodule
get_parameter
get_buffer
get_extra_state
set_extra_state
apply
cuda
ipu
xpu
cpu
type
float
double
half
bfloat16
to_empty
to
register_full_backward_pre_hook
register_backward_hook
register_full_backward_hook
register_forward_pre_hook
register_forward_hook
register_state_dict_pre_hook
state_dict
register_load_state_dict_post_hook
load_state_dict
parameters
named_parameters
buffers
named_buffers
children
named_children
modules
named_modules
train
eval
requires_grad_
zero_grad
share_memory
extra_repr
compile