what does it mean when kernel depth=1 in conv3d pytorch

Question

I would like to understand the difference between conv2d and conv3d in PyTorch.

What is the difference between:

conv3d(in, out, kernel_size(1,3,3))