def test_variance_gaussian_rbm_energy():
new_variance_gaussian_rbm = gaussian_rbm.VarianceGaussianRBM()
samples = torch.ones(1, 128)
energy = new_variance_gaussian_rbm.energy(samples)
assert energy.detach().numpy() < 0
def test_variance_gaussian_rbm_sigma_setter():
new_variance_gaussian_rbm = gaussian_rbm.VarianceGaussianRBM()
try:
new_variance_gaussian_rbm.sigma = 1
except:
new_variance_gaussian_rbm.sigma = torch.nn.Parameter(torch.zeros(128))
assert new_variance_gaussian_rbm.sigma.size(0) == 128
def test_variance_gaussian_rbm_hidden_sampling():
new_variance_gaussian_rbm = gaussian_rbm.VarianceGaussianRBM()
v = torch.ones(1, 128)
probs, states = new_variance_gaussian_rbm.hidden_sampling(v, scale=True)
assert probs.size(1) == 128
assert states.size(1) == 128
probs, states = new_variance_gaussian_rbm.hidden_sampling(v, scale=False)
assert probs.size(1) == 128
assert states.size(1) == 128
def test_variance_gaussian_rbm_visible_sampling():
new_variance_gaussian_rbm = gaussian_rbm.VarianceGaussianRBM()
h = torch.ones(1, 128)
probs, states = new_variance_gaussian_rbm.visible_sampling(h, scale=True)
assert probs.size(1) == 128
assert states.size(1) == 128
probs, states = new_variance_gaussian_rbm.visible_sampling(h, scale=False)
assert probs.size(1) == 128
assert states.size(1) == 128
def test_variance_gaussian_rbm_sigma():
new_variance_gaussian_rbm = gaussian_rbm.VarianceGaussianRBM()
assert new_variance_gaussian_rbm.sigma.size(0) == 128