e78b409a087ed1208f07ef05513eeac9b37f042f,memcnn/models/revop.py,ReversibleBlockFunction,forward,#,123

Before Change

            del x2
            output = torch.cat([y1, y2], dim=1)
            y1.set_()
            y2.set_()
            del y1, y2

        // save the (empty) input and (non-empty) output variables
        ctx.save_for_backward(x, output)

After Change

        with set_grad_enabled(False):
            // partition in two equally sized set of channels
            y1, y2 = torch.chunk(y, 2, dim=1)
            y1, y2 = y1.contiguous(), y2.contiguous()

            // compute outputs
            with warnings.catch_warnings():
                y1var = Variable(y1, requires_grad=False, volatile=True)
            gmr = Gm.forward(y1var).data

            x2 = y2 - gmr
            y2.set_()
            del y2
            with warnings.catch_warnings():
                x2var = Variable(x2, requires_grad=False, volatile=True)
            fmr = Fm.forward(x2var).data
            x1 = y1 - fmr
            y1.set_()
            del y1
            output = torch.cat([x1, x2], dim=1)
            x1.set_()
            x2.set_()
            del x1, x2

Instances