Why do we need the nn.Flatten() layer? Why not going directly to nn.Linear ?

[DimOnGit28]

I am still learning and created some CNN models for different classification use cases. I love Daniel's way of coding compared to others but the one thing I don't understand is why do we need the flatten layer instead of directly going to nn.Linear() as others do?

[DimOnGit28]

I found the answer myself. For those who have the same question :
With the flatten layer you don't need to change the shape of X manually ( x=x.view(-1,10x13*13) )