What happens when the stride is larger than kernel_size?

In Pytorch’s nn.ConvTranspose2d module, we can set the stride to be larger than the kernel_size.
For example:

  • Input matrix of size 2x2
  • Kernel of size 2x2 (kernel_size=2)
  • Stride of 4 (stride=4).

The output image will have dimensions 6x6.
What is going on here?