tensor size in Custom Kernel example

[nrigol]

I am familiarizing with gpytorch and I run the custom kernel example https://docs.gpytorch.ai/en/stable/examples/00_Basic_Usage/Implementing_a_custom_Kernel.html, if I add a print(torch.sin(diff).div(diff)).shape) before returning return torch.sin(diff).div(diff) the outcome is :

1 torch.Size([100, 100])
2 torch.Size([100, 100])
... ....
48 torch.Size([100, 100])
49 torch.Size([100, 100])
50 torch.Size([100, 100])
51 torch.Size([51, 151])
52 torch.Size([100, 100])

I do not understand what is happening to the tensor shape ?
Thank you in advance for your explanation