def test_inheritance(self):
#check if bayesian linear has nn.Module and BayesianModule classes
bconv = BayesianConv3d(in_channels=3,
out_channels=3,
kernel_size=(3, 3, 3),
bias=False)
self.assertEqual(isinstance(bconv, (nn.Module)), True)
self.assertEqual(isinstance(bconv, (BayesianModule)), True)
def test_kl_divergence(self):
#create model, sample weights
#check if kl divergence between apriori and a posteriori is working
bconv = BayesianConv3d(in_channels=3,
out_channels=3,
kernel_size=(3, 3, 3))
to_feed = torch.ones((1, 3, 25, 25, 25))
predicted = bconv(to_feed)
complexity_cost = bconv.log_variational_posterior - bconv.log_prior
self.assertEqual((complexity_cost == complexity_cost).all(),
torch.tensor(True))
pass
def test_variational_inference(self):
#create module, check if inference is variating
bconv = BayesianConv3d(in_channels=3,
out_channels=3,
kernel_size=(3, 3, 3))
conv = nn.Conv3d(in_channels=3, out_channels=3, kernel_size=(3, 3, 3))
to_feed = torch.ones((1, 3, 25, 25, 25))
self.assertEqual((bconv(to_feed) != bconv(to_feed)).any(),
torch.tensor(True))
self.assertEqual((conv(to_feed) == conv(to_feed)).all(),
torch.tensor(True))
pass
def test_weights_shape(self):
#check if weights shape is the expected
bconv = BayesianConv3d(in_channels=3,
out_channels=3,
kernel_size=(3, 3, 3))
conv = nn.Conv3d(in_channels=3, out_channels=3, kernel_size=(3, 3, 3))
to_feed = torch.ones((1, 3, 25, 25, 25))
infer1 = bconv(to_feed)
infer2 = conv(to_feed)
self.assertEqual(infer1.shape, infer2.shape)
pass
def test_weights_shape_cuda(self):
#check if weights shape is the expected
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
bconv = BayesianConv3d(in_channels=3,
out_channels=3,
kernel_size=(3, 3, 3)).to(device)
conv = nn.Conv3d(in_channels=3, out_channels=3,
kernel_size=(3, 3, 3)).to(device)
to_feed = torch.ones((1, 3, 25, 25, 25)).to(device)
infer1 = bconv(to_feed)
infer2 = conv(to_feed)
self.assertEqual(infer1.shape, infer2.shape)
pass
def test_freeze_module(self):
#create module, freeze
#check if two inferences keep equal
bconv = BayesianConv3d(in_channels=3,
out_channels=3,
kernel_size=(3, 3, 3),
bias=False)
to_feed = torch.ones((1, 3, 25, 25, 25))
self.assertEqual((bconv(to_feed) != bconv(to_feed)).any(),
torch.tensor(True))
frozen_feedforward = bconv.forward_frozen(to_feed)
bconv.freeze = True
self.assertEqual((bconv.forward(to_feed) == frozen_feedforward).all(),
torch.tensor(True))
pass
def test_init_bayesian_layer(self):
#create bayesian layer
module = BayesianConv3d(3, 10, (3, 3, 3))
pass