core.layer.ProbConv2d

View Source

 1import torch.nn.functional as f
 2from torch import Tensor, nn
 3
 4from core.layer import AbstractProbLayer
 5
 6
 7class ProbConv2d(nn.Conv2d, AbstractProbLayer):
 8    """
 9    A probabilistic 2D convolution layer.
10
11    Inherits from `nn.Conv2d` and `AbstractProbLayer`. Weights and bias
12    are sampled from associated distributions during forward passes.
13    """
14
15    def forward(self, input: Tensor) -> Tensor:
16        """
17        Perform a 2D convolution using sampled weights and bias.
18
19        Args:
20            input (Tensor): The input tensor of shape (N, C_in, H_in, W_in).
21
22        Returns:
23            Tensor: The output tensor of shape (N, C_out, H_out, W_out).
24        """
25        sampled_weight, sampled_bias = self.sample_from_distribution()
26        return f.conv2d(
27            input,
28            sampled_weight,
29            sampled_bias,
30            self.stride,
31            self.padding,
32            self.dilation,
33            self.groups,
34        )

class ProbConv2d(torch.nn.modules.conv.Conv2d, core.layer.AbstractProbLayer.AbstractProbLayer): View Source

 8class ProbConv2d(nn.Conv2d, AbstractProbLayer):
 9    """
10    A probabilistic 2D convolution layer.
11
12    Inherits from `nn.Conv2d` and `AbstractProbLayer`. Weights and bias
13    are sampled from associated distributions during forward passes.
14    """
15
16    def forward(self, input: Tensor) -> Tensor:
17        """
18        Perform a 2D convolution using sampled weights and bias.
19
20        Args:
21            input (Tensor): The input tensor of shape (N, C_in, H_in, W_in).
22
23        Returns:
24            Tensor: The output tensor of shape (N, C_out, H_out, W_out).
25        """
26        sampled_weight, sampled_bias = self.sample_from_distribution()
27        return f.conv2d(
28            input,
29            sampled_weight,
30            sampled_bias,
31            self.stride,
32            self.padding,
33            self.dilation,
34            self.groups,
35        )

A probabilistic 2D convolution layer.

Inherits from nn.Conv2d and AbstractProbLayer. Weights and bias are sampled from associated distributions during forward passes.

def forward(self, input: torch.Tensor) -> torch.Tensor: View Source

16    def forward(self, input: Tensor) -> Tensor:
17        """
18        Perform a 2D convolution using sampled weights and bias.
19
20        Args:
21            input (Tensor): The input tensor of shape (N, C_in, H_in, W_in).
22
23        Returns:
24            Tensor: The output tensor of shape (N, C_out, H_out, W_out).
25        """
26        sampled_weight, sampled_bias = self.sample_from_distribution()
27        return f.conv2d(
28            input,
29            sampled_weight,
30            sampled_bias,
31            self.stride,
32            self.padding,
33            self.dilation,
34            self.groups,
35        )

Perform a 2D convolution using sampled weights and bias.

Arguments:

input (Tensor): The input tensor of shape (N, C_in, H_in, W_in).

Returns:

Tensor: The output tensor of shape (N, C_out, H_out, W_out).

Inherited Members

core.layer.AbstractProbLayer.AbstractProbLayer: probabilistic_mode; probabilistic; sample_from_distribution