Skip to content

fl_server_ai.tests.test_uncertainty_mc

Classes:

Name Description
MCDropoutTest

Classes

MCDropoutTest

Bases: TestCase


              flowchart TD
              fl_server_ai.tests.test_uncertainty_mc.MCDropoutTest[MCDropoutTest]

              

              click fl_server_ai.tests.test_uncertainty_mc.MCDropoutTest href "" "fl_server_ai.tests.test_uncertainty_mc.MCDropoutTest"

Methods:

Name Description
test_prediction_normal_model
test_prediction_torchscript_model
test_trainer_class
test_trainer_type
Source code in fl_server_ai/tests/test_uncertainty_mc.py 
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
class MCDropoutTest(TestCase):

    def test_trainer_class(self):
        training = Dummy.create_training(uncertainty_method=UncertaintyMethod.MC_DROPOUT)
        trainer_cls = get_trainer_class(training)
        self.assertTrue(trainer_cls is ModelTrainer)

    def test_trainer_type(self):
        training = Dummy.create_training(uncertainty_method=UncertaintyMethod.MC_DROPOUT)
        trainer = get_trainer(training)
        self.assertTrue(type(trainer) is ModelTrainer)

    def test_prediction_torchscript_model(self):
        torch.manual_seed(42)
        layer = torch.nn.Linear(1, 1)
        layer.weight = torch.nn.Parameter(torch.tensor([[1.]]))
        layer.bias = torch.nn.Parameter(torch.tensor([0.]))
        model_bytes = from_torch_module(torch.jit.script(
            torch.nn.Sequential(layer, torch.nn.Dropout(0.5), torch.nn.Tanh())
        ))
        model = Dummy.create_model(GlobalModel, weights=model_bytes)
        Dummy.create_training(
            model=model,
            uncertainty_method=UncertaintyMethod.MC_DROPOUT
        )
        X = torch.tensor([[-4.0], [-2.0], [2.0], [4.0]])
        y = torch.tensor([-1.0, -1.0, 1.0, 1.0])
        logits, uncertainty_dict = MCDropout.prediction(X, model)
        torch.testing.assert_close(y, torch.sign(torch.squeeze(logits)))
        # torch.testing.assert_close(torch.tensor([[0.2667], [0.2330], [0.2663], [0.2333]]),
        #                            uncertainty_dict["variance"], atol=1e-4, rtol=0.001)
        # torch.testing.assert_close(torch.tensor([[0.5164], [0.4827], [0.5161], [0.4830]]),
        #                            uncertainty_dict["std"], atol=1e-4, rtol=0.001)
        self.assertFalse("predictive_entropy" in uncertainty_dict)
        self.assertFalse("expected_entropy" in uncertainty_dict)
        self.assertFalse("mutual_info" in uncertainty_dict)

    def test_prediction_normal_model(self):
        torch.manual_seed(42)
        layer = torch.nn.Linear(1, 1)
        layer.weight = torch.nn.Parameter(torch.tensor([[1.]]))
        layer.bias = torch.nn.Parameter(torch.tensor([0.]))
        model_bytes = from_torch_module(
            torch.nn.Sequential(layer, torch.nn.Dropout(0.5), torch.nn.Tanh())
        )
        model = Dummy.create_model(GlobalModel, weights=model_bytes)
        Dummy.create_training(
            model=model,
            uncertainty_method=UncertaintyMethod.MC_DROPOUT
        )
        X = torch.tensor([[-4.0], [-2.0], [2.0], [4.0]])
        y = torch.tensor([-1.0, -1.0, 1.0, 1.0])
        logits, uncertainty_dict = MCDropout.prediction(X, model)
        torch.testing.assert_close(y, torch.sign(torch.squeeze(logits)))
        # torch.testing.assert_close(torch.tensor([[0.2667], [0.2330], [0.2663], [0.2333]]),
        #                            uncertainty_dict["variance"], atol=1e-4, rtol=0.001)
        # torch.testing.assert_close(torch.tensor([[0.5164], [0.4827], [0.5161], [0.4830]]),
        #                            uncertainty_dict["std"], atol=1e-4, rtol=0.001)
        self.assertFalse("predictive_entropy" in uncertainty_dict)
        self.assertFalse("expected_entropy" in uncertainty_dict)
        self.assertFalse("mutual_info" in uncertainty_dict)

Functions

test_prediction_normal_model 
test_prediction_normal_model()
Source code in fl_server_ai/tests/test_uncertainty_mc.py 
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
def test_prediction_normal_model(self):
    torch.manual_seed(42)
    layer = torch.nn.Linear(1, 1)
    layer.weight = torch.nn.Parameter(torch.tensor([[1.]]))
    layer.bias = torch.nn.Parameter(torch.tensor([0.]))
    model_bytes = from_torch_module(
        torch.nn.Sequential(layer, torch.nn.Dropout(0.5), torch.nn.Tanh())
    )
    model = Dummy.create_model(GlobalModel, weights=model_bytes)
    Dummy.create_training(
        model=model,
        uncertainty_method=UncertaintyMethod.MC_DROPOUT
    )
    X = torch.tensor([[-4.0], [-2.0], [2.0], [4.0]])
    y = torch.tensor([-1.0, -1.0, 1.0, 1.0])
    logits, uncertainty_dict = MCDropout.prediction(X, model)
    torch.testing.assert_close(y, torch.sign(torch.squeeze(logits)))
    # torch.testing.assert_close(torch.tensor([[0.2667], [0.2330], [0.2663], [0.2333]]),
    #                            uncertainty_dict["variance"], atol=1e-4, rtol=0.001)
    # torch.testing.assert_close(torch.tensor([[0.5164], [0.4827], [0.5161], [0.4830]]),
    #                            uncertainty_dict["std"], atol=1e-4, rtol=0.001)
    self.assertFalse("predictive_entropy" in uncertainty_dict)
    self.assertFalse("expected_entropy" in uncertainty_dict)
    self.assertFalse("mutual_info" in uncertainty_dict)
test_prediction_torchscript_model 
test_prediction_torchscript_model()
Source code in fl_server_ai/tests/test_uncertainty_mc.py 
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
def test_prediction_torchscript_model(self):
    torch.manual_seed(42)
    layer = torch.nn.Linear(1, 1)
    layer.weight = torch.nn.Parameter(torch.tensor([[1.]]))
    layer.bias = torch.nn.Parameter(torch.tensor([0.]))
    model_bytes = from_torch_module(torch.jit.script(
        torch.nn.Sequential(layer, torch.nn.Dropout(0.5), torch.nn.Tanh())
    ))
    model = Dummy.create_model(GlobalModel, weights=model_bytes)
    Dummy.create_training(
        model=model,
        uncertainty_method=UncertaintyMethod.MC_DROPOUT
    )
    X = torch.tensor([[-4.0], [-2.0], [2.0], [4.0]])
    y = torch.tensor([-1.0, -1.0, 1.0, 1.0])
    logits, uncertainty_dict = MCDropout.prediction(X, model)
    torch.testing.assert_close(y, torch.sign(torch.squeeze(logits)))
    # torch.testing.assert_close(torch.tensor([[0.2667], [0.2330], [0.2663], [0.2333]]),
    #                            uncertainty_dict["variance"], atol=1e-4, rtol=0.001)
    # torch.testing.assert_close(torch.tensor([[0.5164], [0.4827], [0.5161], [0.4830]]),
    #                            uncertainty_dict["std"], atol=1e-4, rtol=0.001)
    self.assertFalse("predictive_entropy" in uncertainty_dict)
    self.assertFalse("expected_entropy" in uncertainty_dict)
    self.assertFalse("mutual_info" in uncertainty_dict)
test_trainer_class 
test_trainer_class()
Source code in fl_server_ai/tests/test_uncertainty_mc.py 
18
19
20
21
def test_trainer_class(self):
    training = Dummy.create_training(uncertainty_method=UncertaintyMethod.MC_DROPOUT)
    trainer_cls = get_trainer_class(training)
    self.assertTrue(trainer_cls is ModelTrainer)
test_trainer_type 
test_trainer_type()
Source code in fl_server_ai/tests/test_uncertainty_mc.py 
23
24
25
26
def test_trainer_type(self):
    training = Dummy.create_training(uncertainty_method=UncertaintyMethod.MC_DROPOUT)
    trainer = get_trainer(training)
    self.assertTrue(type(trainer) is ModelTrainer)

Functions