def setUp(self):
self.dim = int(10 + torch.randint(100, (1, 1)).item())
self.dim = 2
self.conf_files = []
for path in glob.glob('./tests/models/*yml'):
self.conf_files.append(path)
assert len(self.conf_files) > 0
self.mean = torch.randn(self.dim).type(self.type)
self.variance = (1 + torch.randn(self.dim)**2).type(self.type)
self.models1 = []
for conf_file in self.conf_files:
with open(conf_file, 'r') as fid:
data = fid.read().replace('<feadim>', str(self.dim))
conf = yaml.load(data)
model = beer.create_model(conf, self.mean, self.variance)
self.models1.append(model)
self.mean = torch.randn(self.dim).type(self.type)
self.variance = (1 + torch.randn(self.dim)**2).type(self.type)
self.models2 = []
for conf_file in self.conf_files:
with open(conf_file, 'r') as fid:
data = fid.read().replace('<feadim>', str(self.dim))
conf = yaml.load(data)
model = beer.create_model(conf, self.mean, self.variance)
self.models2.append(model)
self.weights = (1 + torch.randn(2)**2).type(self.type)
self.weights /= self.weights.sum()
def setUp(self):
self.device = 'cpu'
self.dim = int(10 + torch.randint(100, (1, 1)).item())
self.npoints = int(1 + torch.randint(100, (1, 1)).item())
self.data = torch.randn(self.npoints, self.dim).type(self.type)
self.mean = torch.randn(self.dim).type(self.type)
self.variance = (1 + torch.randn(self.dim)**2).type(self.type)
self.conf_files = []
for path in glob.glob('./tests/models/*yml'):
if not os.path.basename(path).startswith('normalset') and \
not os.path.basename(path).startswith('bernoulli'):
self.conf_files.append(path)
assert len(self.conf_files) > 0
self.models = []
for conf_file in self.conf_files:
with open(conf_file, 'r') as fid:
data = fid.read()
data = data.replace('<feadim>', str(self.dim))
conf = yaml.load(data)
model = beer.create_model(conf, self.mean, self.variance)
self.models.append(model)
self.acc_stats1 = {
'shared_key': torch.randn(self.dim).type(self.type),
'key1': torch.randn(self.dim).type(self.type),
}
self.acc_stats2 = {
'shared_key': torch.randn(self.dim).type(self.type),
'key2': torch.randn(self.dim).type(self.type),
}
def setUp(self):
self.dim = int(1 + torch.randint(100, (1, 1)).item())
self.npoints = int(1 + torch.randint(100, (1, 1)).item())
self.mean = torch.randn(self.dim).type(self.type)
self.variance = 1 + torch.randn(self.dim).type(self.type) ** 2
self.pseudo_counts = 1e-2 + 100 * torch.rand(1).item()
self.data = torch.randn(self.npoints, self.dim).type(self.type)
self.means = torch.randn(self.npoints, self.dim).type(self.type)
self.vars = torch.randn(self.npoints, self.dim).type(self.type) ** 2
with open('./tests/models/normal_iso.yml') as fid:
conf = yaml.load(fid)
self.model = beer.create_model(conf, self.mean, self.variance)
def test_create_model(self):
for conf_file in self.conf_files:
with self.subTest(conf_file=conf_file):
# Doesn't test the content of the object, just make sure
# that the creation does not crash and return a valid
# object.
with open(conf_file, 'r') as fid:
data = fid.read()
data = data.replace('<feadim>', str(self.dim))
conf = yaml.load(data)
model = beer.create_model(conf, self.mean, self.variance)
self.assertTrue(
isinstance(model, beer.BayesianModel)
or isinstance(model, torch.nn.Module))
def setUp(self):
self.dim = int(1 + torch.randint(100, (1, 1)).item())
self.npoints = int(1 + torch.randint(100, (1, 1)).item())
self.data = torch.randn(self.npoints, self.dim).type(self.type)
self.means = torch.randn(self.npoints, self.dim).type(self.type)
self.vars = torch.randn(self.npoints, self.dim).type(self.type) ** 2
self.mean = torch.randn(self.dim).type(self.type)
self.variance = 1 + torch.randn(self.dim).type(self.type) ** 2
cov = (1 + torch.randn(self.dim)).type(self.type)
self.cov = torch.eye(self.dim).type(self.type) + torch.ger(cov, cov)
self.pseudo_counts = 1e-2 + 100 * torch.rand(1).item()
self.ncomp = int(1 + torch.randint(100, (1, 1)).item())
self.prior_mean = torch.randn(self.dim).type(self.type)
with open('./tests/models/normalset_full_sharedcov.yml') as fid:
conf = yaml.load(fid)
self.model = beer.create_model(conf, self.mean, self.variance)
def setUp(self):
self.dim = int(10 + torch.randint(100, (1, 1)).item())
self.dim_subspace = int(1 + torch.randint(self.dim - 1, (1, 1)).item())
self.npoints = int(1 + torch.randint(100, (1, 1)).item())
self.data = torch.randn(self.npoints, self.dim).type(self.type)
self.latent = torch.randn(self.npoints,
self.dim_subspace).type(self.type)
self.means = torch.randn(self.npoints, self.dim).type(self.type)
self.vars = torch.randn(self.npoints, self.dim).type(self.type)**2
# Create the PPCA model
self.mean = torch.randn(self.dim).type(self.type)
self.variance = 1 + torch.randn(self.dim).type(self.type)**2
with open('./tests/models/ppca.yml') as fid:
conf = yaml.load(fid)
self.model = beer.create_model(conf, self.mean, self.variance)
self.dim_subspace = self.model.subspace.shape[0]
def setUp(self):
self.npoints = int(1 + torch.randint(100, (1, 1)).item())
self.dim = int(1 + torch.randint(100, (1, 1)).item())
self.data = torch.randn(self.npoints, self.dim).type(self.type)
self.mean = torch.randn(self.dim).type(self.type)
self.variance = 1 + torch.randn(self.dim).type(self.type) ** 2
self.means = torch.randn(self.npoints, self.dim).type(self.type)
self.vars = torch.randn(self.npoints, self.dim).type(self.type) ** 2
self.pseudo_counts = 1e-2 + 100 * torch.rand(1).item()
self.ncomp = int(1 + torch.randint(100, (1, 1)).item())
self.weights = (1 + torch.randn(self.ncomp) ** 2).type(self.type)
self.weights /= self.weights.sum()
self.mixtures = []
for path in glob.glob('./tests/models/*yaml'):
if os.path.basename(path).startswith('mixture'):
with open(path) as fid:
conf = yaml.load(fid)
model = beer.create_model(conf, self.mean, self.variance)
self.mixtures.append(model)
def setUp(self):
self.dim = int(10 + torch.randint(100, (1, 1)).item())
self.nclasses = int(1 + torch.randint(20, (1, 1)).item())
self.dim_subspace1 = int(1 +
torch.randint(self.dim - 1, (1, 1)).item())
self.dim_subspace2 = int(1 +
torch.randint(self.dim - 1, (1, 1)).item())
self.npoints = int(1 + torch.randint(100, (1, 1)).item())
self.data = torch.randn(self.npoints, self.dim).type(self.type)
self.means = torch.randn(self.npoints, self.dim).type(self.type)
self.vars = torch.randn(self.npoints, self.dim).type(self.type)**2
# Create the PLDA model
self.mean = torch.randn(self.dim).type(self.type)
self.variance = 1 + torch.randn(self.dim).type(self.type)**2
with open('./tests/models/plda.yml') as fid:
conf = yaml.load(fid)
self.model = beer.create_model(conf, self.mean, self.variance).modelset
self.dim_subspace1 = self.model.noise_subspace.shape[0]
self.dim_subspace2 = self.model.class_subspace.shape[0]
