Thanks alot.
I have tried something like:
model = resnet18()
model.conv1 = nn.Conv2d(4, 64, kernel_size=7, stride=2, padding=3, bias=False)
from this post Transfer learning with different inputs - #5 by austin
1 Like