The input of branch_net is set to requires_grad

[luoyh15]

I notice that the input of branch_net is set to requires_grad in the deepxde. Is it nesseary?
As I understand, the input of branch_net's grad is not used.
The result should not be affected, while maybe slow down the training step.

The piece of code is from deepxde/model.py 305-317, and shows the setting step.

        def outputs(training, inputs):
            self.net.train(mode=training)
            with torch.no_grad():
                if isinstance(inputs, tuple):
                    inputs = tuple(
                        map(lambda x: torch.as_tensor(x).requires_grad_(), inputs)
                    )
                else:
                    inputs = torch.as_tensor(inputs)
                    inputs.requires_grad_()
            # Clear cached Jacobians and Hessians.
            grad.clear()
            return self.net(inputs)