def test_lkj_prior_log_prob(self, cuda=False):
device = torch.device("cuda") if cuda else torch.device("cpu")
prior = LKJPrior(2, torch.tensor(0.5, device=device))
S = torch.eye(2, device=device)
self.assertAlmostEqual(prior.log_prob(S).item(), -1.86942, places=4)
S = torch.stack([S, torch.tensor([[1.0, 0.5], [0.5, 1]], device=S.device)])
self.assertTrue(approx_equal(prior.log_prob(S), torch.tensor([-1.86942, -1.72558], device=S.device)))
with self.assertRaises(ValueError):
prior.log_prob(torch.eye(3, device=device))
# For eta=1.0 log_prob is flat over all covariance matrices
prior = LKJPrior(2, torch.tensor(1.0, device=device))
self.assertTrue(torch.all(prior.log_prob(S) == prior.C))
def test_lkj_prior_batch_log_prob(self, cuda=False):
device = torch.device("cuda") if cuda else torch.device("cpu")
prior = LKJPrior(2, torch.tensor([0.5, 1.5], device=device))
S = torch.eye(2, device=device)
self.assertTrue(approx_equal(prior.log_prob(S), torch.tensor([-1.86942, -0.483129], device=S.device)))
S = torch.stack([S, torch.tensor([[1.0, 0.5], [0.5, 1]], device=S.device)])
self.assertTrue(approx_equal(prior.log_prob(S), torch.tensor([-1.86942, -0.62697], device=S.device)))
with self.assertRaises(ValueError):
prior.log_prob(torch.eye(3, device=device))
def test_lkj_prior_sample(self, seed=0):
torch.random.manual_seed(seed)
prior = LKJPrior(n=5, eta=0.5)
random_samples = prior.sample(torch.Size((8, )))
self.assertTrue(_is_valid_correlation_matrix(random_samples))
max_non_symm = (random_samples -
random_samples.transpose(-1, -2)).abs().max()
self.assertLess(max_non_symm, 1e-4)
self.assertEqual(random_samples.shape, torch.Size((8, 5, 5)))
def test_lkj_prior_batch_log_prob(self, cuda=False):
device = torch.device("cuda") if cuda else torch.device("cpu")
prior = LKJPrior(2, torch.tensor([0.5, 1.5], device=device))
dist = LKJCholesky(2, torch.tensor([0.5, 1.5], device=device))
S = torch.eye(2, device=device)
S_chol = torch.linalg.cholesky(S)
self.assertTrue(approx_equal(prior.log_prob(S), dist.log_prob(S_chol)))
S = torch.stack(
[S, torch.tensor([[1.0, 0.5], [0.5, 1]], device=S.device)])
S_chol = torch.linalg.cholesky(S)
self.assertTrue(approx_equal(prior.log_prob(S), dist.log_prob(S_chol)))
with self.assertRaises(ValueError):
prior.log_prob(torch.eye(3, device=device))
def test_lkj_prior_validate_args(self):
LKJPrior(2, 1.0, validate_args=True)
with self.assertRaises(ValueError):
LKJPrior(1.5, 1.0, validate_args=True)
with self.assertRaises(ValueError):
LKJPrior(2, -1.0, validate_args=True)
def test_lkj_prior_to_gpu(self):
if torch.cuda.is_available():
prior = LKJPrior(2, 1.0).cuda()
self.assertEqual(prior.eta.device.type, "cuda")
self.assertEqual(prior.C.device.type, "cuda")