RuntimeError: Given groups=1, weight of size [64, 164, 3, 3], expected input[8, 64, 256, 256] to have 164 channels, but got 64 channels instead

i am trying to create a generator with depthwise, feature reused residual block and inception block but the getting an error i have printed the shapes of the previous layers as follows

DepthwiseDilatedConv latent_noise - Output Shape: torch.Size([8, 100, 256, 256])
DepthwiseDilatedConv torch.cat - Output Shape: torch.Size([8, 101, 256, 256])
DepthwiseDilatedConv self depthwise_conv - Output Shape: torch.Size([8, 101, 256, 256])
DepthwiseDilatedConv self pointwise_conv - Output Shape: torch.Size([8, 64, 256, 256])
FeatureReuseResidualBlock 'residual' - Output Shape: torch.Size([8, 64, 256, 256])

but this line of code is giving an error

x = self.conv1(x)

this is the full code

import torch
import torch.nn as nn

# Depthwise dilated convolutional layer
class DepthwiseDilatedConv(nn.Module):
    def __init__(self, in_channels, latent_dim, out_channels, dilation_factor):
        super(DepthwiseDilatedConv, self).__init__()
        self.depthwise_conv = nn.Conv2d(in_channels + latent_dim, in_channels + latent_dim, kernel_size=3, padding=dilation_factor, dilation=dilation_factor, groups=in_channels + latent_dim)
        self.pointwise_conv = nn.Conv2d(in_channels + latent_dim, out_channels, kernel_size=1)

    def forward(self, x, latent_noise):
        # Ensure that the dimensions are compatible for concatenation
        latent_noise = latent_noise.expand(-1, -1, x.size(2), x.size(3))
        print("DepthwiseDilatedConv latent_noise - Output Shape:", latent_noise.shape)
        # Concatenate latent noise along the channel dimension
        x = torch.cat([x, latent_noise], dim=1)
        print("DepthwiseDilatedConv torch.cat - Output Shape:", x.shape)
        x = self.depthwise_conv(x)
        print("DepthwiseDilatedConv self depthwise_conv - Output Shape:", x.shape)
        x = self.pointwise_conv(x)
        print("DepthwiseDilatedConv self pointwise_conv - Output Shape:", x.shape)
        return x

# Feature Reuse Residual Block
class FeatureReuseResidualBlock(nn.Module):
    def __init__(self, in_channels, out_channels):
        super(FeatureReuseResidualBlock, self).__init__()
        self.conv1 = nn.Conv2d(in_channels, out_channels, kernel_size=3, padding=1)
        self.conv2 = nn.Conv2d(out_channels, in_channels, kernel_size=3, padding=1)

    def forward(self, x):
        residual = x
        print("FeatureReuseResidualBlock 'residual' - Output Shape:", residual.shape)
        x = self.conv1(x)
        x = nn.functional.relu(x, inplace=True)
        x = self.conv2(x)
        residual = nn.functional.pad(residual, (0, 0, 0, 0, 0, x.shape[1] - residual.shape[1]))
        x += residual
        x = nn.functional.relu(x, inplace=True)
        print("FeatureReuseResidualBlock - Output Shape:", x.shape)
        return x

# InceptionV3 with Factorization
class InceptionV3Factorized(nn.Module):
    def __init__(self, in_channels, out_channels):
        super(InceptionV3Factorized, self).__init__()
        self.conv1x1 = nn.Conv2d(in_channels, out_channels // 2, kernel_size=1)

        self.conv3x3_reduce = nn.Conv2d(in_channels, out_channels // 4, kernel_size=1)
        self.conv3x3_1 = nn.Conv2d(out_channels // 4, out_channels // 4, kernel_size=(1, 3), padding=(0, 1))
        self.conv3x3_2 = nn.Conv2d(out_channels // 4, out_channels // 2, kernel_size=(3, 1), padding=(1, 0))

        self.conv5x5_reduce = nn.Conv2d(in_channels, out_channels // 4, kernel_size=1)
        self.conv5x5_1 = nn.Conv2d(out_channels // 4, out_channels // 4, kernel_size=(1, 5), padding=(0, 2))
        self.conv5x5_2 = nn.Conv2d(out_channels // 4, out_channels // 2, kernel_size=(5, 1), padding=(2, 0))

        self.pool = nn.MaxPool2d(kernel_size=3, stride=1, padding=1)
        self.pool_proj = nn.Conv2d(in_channels, out_channels // 2, kernel_size=1)

    def forward(self, x):
        x1 = nn.functional.relu(self.conv1x1(x), inplace=True)
        x2 = nn.functional.relu(self.conv3x3_reduce(x), inplace=True)
        x2_1 = nn.functional.relu(self.conv3x3_1(x2), inplace=True)
        x2_2 = nn.functional.relu(self.conv3x3_2(x2), inplace=True)
        x3 = nn.functional.relu(self.conv5x5_reduce(x), inplace=True)
        x3_1 = nn.functional.relu(self.conv5x5_1(x3), inplace=True)
        x3_2 = nn.functional.relu(self.conv5x5_2(x3), inplace=True)
        x4 = nn.functional.relu(self.pool(x))
        x4 = nn.functional.relu(self.pool_proj(x4), inplace=True)
        output = torch.cat([x1, x2_1, x2_2, x3_1, x3_2, x4], dim=1)
        print("InceptionV3Factorized - Output Shape:", output.shape)
        return output

# GAN Generator Architecture
class Generator(nn.Module):
    def __init__(self, image_channels, latent_dim, channels=64):
        super(Generator, self).__init__()

        self.depthwise_dilated_conv = DepthwiseDilatedConv(image_channels, latent_dim, channels, dilation_factor=2)
        self.feature_reuse_residual_block = FeatureReuseResidualBlock(channels + latent_dim, channels)
        self.inception_block = InceptionV3Factorized(channels, image_channels)

        self.deconv1 = nn.ConvTranspose2d(channels + latent_dim, 128, kernel_size=4, stride=2, padding=1)
        self.deconv2 = nn.ConvTranspose2d(128, 64, kernel_size=4, stride=2, padding=1)
        self.final_conv = nn.Conv2d(64, image_channels, kernel_size=3, stride=1, padding=1)
        self.tanh = nn.Tanh()

    def forward(self, images, latent_noise):
        x = self.depthwise_dilated_conv(images, latent_noise)
        x = self.feature_reuse_residual_block(x)
        x = self.inception_block(x)
        x = nn.functional.relu(self.deconv1(x), inplace=True)
        x = nn.functional.relu(self.deconv2(x), inplace=True)
        x = self.tanh(self.final_conv(x))
        print("Generator - Final Output Shape:", x.shape)
        return x

# Testing the model with random noise and images
torch.manual_seed(42)

# Instantiate the generator
image_channels = 1  # Grayscale images
latent_dim = 100
generator = Generator(image_channels, latent_dim, channels=64)

# Generate random noise and images
random_noise = torch.randn(8, latent_dim, 1, 1)
random_images = torch.randn(8, image_channels, 256, 256)

# Forward pass through the generator
generated_image = generator(random_images, random_noise)

# Print the shape of the generated image
print("Generated Image Shape:", generated_image.shape)

In your Generator model you are setting the output channels of DepthwiseDilatedConv to channels=64 which will create an activation in the shape torch.Size([8, 64, 256, 256]) here:

    def forward(self, images, latent_noise):
        x = self.depthwise_dilated_conv(images, latent_noise)
        print(x.shape) <----

You are then passing this tensor to self.feature_reuse_residual_block(x) which was initialized as:

FeatureReuseResidualBlock(channels + latent_dim)

channels + latent_dim = 164 so this module expects inputs with 164 channels while x has 64. Either change the output channels of the previous layer to 164 or change the input channels to FeatureReuseResidualBlock to 64.

thanks a lot i added latent_dim and it worked but now inception class is given me an error of

RuntimeError: Given groups=1, expected weight to be at least 1 at dimension 0, but got weight of size [0, 164, 1, 1] instead

as i also added the latent_dim to its channel in the generator

self.inception_block = InceptionV3Factorized(channels + latent_dim, image_channels)

this is where the error is on x1

x1 = nn.functional.relu(self.conv1x1(x), inplace=True)

Your out_channel value might be too small creating an empty kernel:

in_channels = 164
out_channels = 3
print(nn.Conv2d(in_channels, out_channels // 4, 3).weight.shape)
#  torch.Size([0, 164, 3, 3])

Please can your clarify i really did not understand.
Thank you

My code snippet shows the issue since 3//4 will calculate 0 which is then used for the output channels.