Lacks a `ff` layer?

[CiaoHe]

Hi Phil!
Compared to Algorithm3 Line 4, I found a little inconsistent,

recurrent-interface-network-pytorch/rin_pytorch/rin_pytorch.py

Line 310 in 1986201

latents = self.latents_attend_to_patches(latents, patches, time = t) + latents

maybe need

latents = self.latents_attend_to_patches(latents, patches, time = t) + latents 
latents = self.latents_cross_attn_ff(latents, time=t) + latents

just curious 😁

[lucidrains]

@CiaoHe Hi He Cao! Yes indeed all attention should be followed by feedforwards, thank you for catching this!