test_torch.py 文件源码

def _test_neg(self, cast):
        float_types = ['torch.DoubleTensor', 'torch.FloatTensor', 'torch.LongTensor']
        int_types = ['torch.IntTensor', 'torch.ShortTensor', 'torch.ByteTensor',
                     'torch.CharTensor']

        for t in float_types + int_types:
            if t in float_types:
                a = cast(torch.randn(100, 90).type(t))
            else:
                a = cast(torch.Tensor(100, 90).type(t).random_())
            zeros = cast(torch.Tensor().type(t)).resize_as_(a).zero_()

            res_add = torch.add(zeros, -1, a)
            res_neg = a.clone()
            res_neg.neg_()
            self.assertEqual(res_neg, res_add)

            # test out of place as well
            res_neg_out_place = a.clone().neg()
            self.assertEqual(res_neg_out_place, res_add)

            # test via __neg__ operator
            res_neg_op = -a.clone()
            self.assertEqual(res_neg_op, res_add)