span sizes

[amhsinyednap]

I have 256x256 images, I read your paper which uses a 65x65 span. But in my case of 256x256 for doing axial attention once, can I use a 256x256 span or I should use the conv-stem to decrease the size to 56 as in your paper and use a span of 56x56?

[csrhddlam]

It's up to your choice.

Conv-stem is faster but might not be as good as applying attention directly.

[amhsinyednap]

So I can use a 256x 256 span if my image size is 256, and no need to decrease its size to 56?

[csrhddlam]

Yes, you can do that.