Untitled

class TransformerBlock(nn.Module):
    # Vision Transformer https://arxiv.org/abs/2010.11929
    def __init__(self, c1, c2, num_heads, num_layers):
        super().__init__()
        self.conv = None
        if c1 != c2:
            self.conv = Conv(c1, c2)
        self.linear = nn.Linear(c2, c2)  # learnable position embedding
        self.tr = nn.Sequential(*(TransformerLayer(c2, num_heads) for _ in range(num_layers)))
        self.c2 = c2
        self.to_patch = Conv(c1 * 5, c1)

    def forward(self, x):
        shifts = ((1, -1, 0, 0), (-1, 1, 0, 0), (0, 0, 1, -1), (0, 0, -1, 1))
        shifted_x = list(map(lambda shift: F.pad(x, shift), shifts))
        x_with_shifts = torch.cat((x, *shifted_x), dim = 1)
        x = self.to_patch(x_with_shifts)

        if self.conv is not None:
            x = self.conv(x)
        b, _, w, h = x.shape
        p = x.flatten(2).permute(2, 0, 1)
        return self.tr(p + self.linear(p)).permute(1, 2, 0).reshape(b, self.c2, w, h)