First layer with single channel

🚀 Feature

When calling the model, it would be nice to be able to set the number of channels in the first layer.

Motivation

When using a grayscale image, I want the input to be a single channel.

Pitch

ex)

self.conv1 = nn.Conv2d(3, self.inplanes, kernel_size=7, stride=2, padding=3, bias=False)
-> self.conv1 = nn.Conv2d(1, self.inplanes, kernel_size=7, stride=2, padding=3, bias=False)

layers.append(ConvBNActivation(3, firstconv_output_channels, kernel_size=3, stride=2, norm_layer=norm_layer, activation_layer=nn.Hardswish))
-> layers.append(ConvBNActivation(1, firstconv_output_channels, kernel_size=3, stride=2, norm_layer=norm_layer,  activation_layer=nn.Hardswish))

Alternatives

My idea is to use parameters when calling the model.
ex)
self.conv1 = nn.Conv2d(input, self.inplanes, kernel_size=7, stride=2, padding=3, bias=False)

Additional context

No files or tests are named; start by locating the model entry points and first convolution definitions referenced by the issue. Check how model-call parameters are handled, then verify that a caller can select one input channel for grayscale images without breaking the existing three-channel behavior.