A mistake in SSPSR.py

[murphyhoucn]

There is a mistake in the code file SSPSR.py, line 81 to 87. The code is below.

        if self.shared:
            self.branch = BranchUnit(n_subs, n_feats, n_blocks, act, res_scale, up_scale=n_scale//2, conv=default_conv)
            # up_scale=n_scale//2 means that we upsample the LR input n_scale//2 at the branch network, and then conduct 2 times upsampleing at the global network
        else:
            self.branch = nn.ModuleList()
            for i in range(self.G):
                self.branch.append(BranchUnit(n_subs, n_feats, n_blocks, act, res_scale, up_scale=2, conv=default_conv))

The parameter in line 87 up_scale=2 should alse be 'up_scale=n_scale//2'. if not, the tensor addition may result in tensor size mismatch.