The Conv2d function only has 4 types of padding mode which are ‘zeros’ , ‘reflect’ , ‘replicate’ or ‘circular’. I want to know how I could do symmetric padding using pytorch. For example, I want a tensor [[1, 2, 3], [4, 5, 6]] to become this one after padding. And I want to know what the circular padding mode is. Could you give me some visulization to explain the circular padding? Really appreciate your answer!
[[2, 1, 1, 2, 3, 3, 2],
[2, 1, 1, 2, 3, 3, 2],
[5, 4, 4, 5, 6, 6, 5],
[5, 4, 4, 5, 6, 6, 5]].