def test_batch_eval_neg_holder_table(self, cuda=False): device = torch.device("cuda") if cuda else torch.device("cpu") for dtype in (torch.float, torch.double): X = torch.zeros(2, 2, device=device, dtype=dtype) res = neg_holder_table(X) self.assertEqual(res.dtype, dtype) self.assertEqual(res.device.type, device.type) self.assertEqual(res.shape, torch.Size([2])) self.assertTrue(res.abs().sum().item() < 1e-6)
def test_neg_holder_table_global_maxima(self, cuda=False): device = torch.device("cuda") if cuda else torch.device("cpu") for dtype in (torch.float, torch.double): X = torch.tensor(GLOBAL_MAXIMIZERS, device=device, dtype=dtype, requires_grad=True) res = neg_holder_table(X) torch.autograd.backward([*res]) self.assertTrue(torch.max((res - GLOBAL_MAXIMUM).abs()) < 1e-5) self.assertLess(X.grad.abs().max().item(), 1e-3)