Apply the patch-merger projection to each variable-sized group

Read the equation

For each input item, normalize its per-patch feature vectors, flatten each patch group while retaining its leading group count, and apply the merger projection. The list form permits different items to have different numbers of patch groups. Its feature-width requirements are the same as the tensor path.

Terms & definitions

The recovered passage does not supply a separate symbol table. Read the source context for the notation used in this version; familiar letters may carry different meanings in other builds.

How it works

Read the definition from its right-hand expression to the quantity on the left. Its surrounding source records the intended input, state, or parameter context. Where the source leaves an operation unspecified, this page preserves that gap.

Assumptions & limits

Source recovery establishes that the expression appears in this record. Its mathematical validity, implementation behavior, and interpretation must be assessed separately. See the linked source and research notes for the checks available for this version.

Original context

def forward(self, x, *args, **kwargs):
        if isinstance(x, list) or isinstance(x, tuple):
            x = [
                self.proj(self.pre_norm(item).view(item.shape[0], -1))
                for item in x
            ]
        else:
            # B, N, N_k, C = x.shape
            B = x.shape[0]

x = [
                self.proj(self.pre_norm(item).view(item.shape[0], -1))
                for item in x
            ]

Related builds