Method forward_raw

models/aios/backbones/swin_transformer.py:670–705

Forward function.

(self, x)

    #         raise TypeError('pretrained must be a str or None')

    def forward_raw(self, x):
        """Forward function."""
        x = self.patch_embed(x)

        Wh, Ww = x.size(2), x.size(3)
        if self.ape:
            # interpolate the position embedding to the corresponding size
            absolute_pos_embed = F.interpolate(self.absolute_pos_embed,
                                               size=(Wh, Ww),
                                               mode='bicubic')
            x = (x + absolute_pos_embed).flatten(2).transpose(1,
                                                              2)  # B Wh*Ww C
        else:
            x = x.flatten(2).transpose(1, 2)
        x = self.pos_drop(x)

        outs = []
        for i in range(self.num_layers):
            layer = self.layers[i]
            x_out, H, W, x, Wh, Ww = layer(x, Wh, Ww)
            # import pdb; pdb.set_trace()

            if i in self.out_indices:
                norm_layer = getattr(self, f'norm{i}')
                x_out = norm_layer(x_out)

                out = x_out.view(-1, H, W,
                                 self.num_features[i]).permute(0, 3, 1,
                                                               2).contiguous()
                outs.append(out)
        # in:
        #   torch.Size([2, 3, 1024, 1024])
        # outs:
        #   [torch.Size([2, 192, 256, 256]), torch.Size([2, 384, 128, 128]), \
        #       torch.Size([2, 768, 64, 64]), torch.Size([2, 1536, 32, 32])]
        return tuple(outs)

    def forward(self, tensor_list: NestedTensor):
        x = tensor_list.tensors