Patch Embedding Design Choice?

[tonyyunyang]

Hi,
I am very curious about why you chose to design the patch embedding function as below:

self.to_patch_embedding = nn.Sequential(
    Rearrange('b c (f pf) (h p1) (w p2) -> b f (h w) (p1 p2 pf c)', p1 = patch_height, p2 = patch_width, pf = frame_patch_size),
    nn.LayerNorm(patch_dim),
    nn.Linear(patch_dim, dim),
    nn.LayerNorm(dim)
)

I have seen people using other methods, like 2DConv, 3DConv, etc.
I am very curious how you came up with this, because it works extremely well.