def forward()

in src/resnet50.py [0:0]

• 16 lines of code
• 5 McCabe index (conditional complexity)

    def forward(self, inputs):
        if not isinstance(inputs, list):
            inputs = [inputs]
        idx_crops = torch.cumsum(torch.unique_consecutive(
            torch.tensor([inp.shape[-1] for inp in inputs]),
            return_counts=True,
        )[1], 0)
        start_idx = 0
        for end_idx in idx_crops:
            _out = self.forward_backbone(torch.cat(inputs[start_idx: end_idx]).cuda(non_blocking=True))
            if start_idx == 0:
                output = _out
            else:
                output = torch.cat((output, _out))
            start_idx = end_idx
        return self.forward_head(output)