-
Notifications
You must be signed in to change notification settings - Fork 0
/
test.py
173 lines (144 loc) · 8.19 KB
/
test.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
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
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
import importlib
import os
import sys
import unittest
from pathlib import Path
import numpy as np
import torch
from torch import nn
from torch.utils.data import TensorDataset, DataLoader
class HidePrints:
"""Disable normal printing."""
def __enter__(self):
self._original_stdout = sys.stdout
sys.stdout = open(os.devnull, 'w')
def __exit__(self, exc_type, exc_val, exc_tb):
sys.stdout.close()
sys.stdout = self._original_stdout
class TestProject(unittest.TestCase):
def __init__(self, *args, **kwargs):
super(TestProject, self).__init__(*args, **kwargs)
self.project_path = Path(".")
self.no_print = HidePrints
@staticmethod
def title(msg):
print(f"\n==============\n> {msg} ...")
def test_1_folder_structure(self):
"""Test the framework structure (folder and files)."""
self.title("Testing folder structure")
self.assertTrue(self.project_path.exists(), f"No folder found at {self.project_path}")
# Main files
for file in ["main.py"]:
with self.subTest(f"Checking file {file}"):
self.assertTrue((self.project_path / file).exists(), f"No file {file} found at {self.project_path}")
# Source code
src_path = self.project_path / "src"
self.assertTrue(src_path.exists(), f"{src_path} not found")
for file in ["__init__.py", "data.py", "utils.py"]:
with self.subTest(f"Checking file src/{file}"):
self.assertTrue((src_path / file).exists(), f"No file {file} found at {src_path}")
# Methods
method_path = src_path / "methods"
self.assertTrue(method_path.exists(), f"{method_path} not found")
for file in ["__init__.py", "dummy_methods.py",
"pca.py", "deep_network.py"]:
with self.subTest(f"Checking file methods/{file}"):
self.assertTrue((method_path / file).exists(), f"No file {file} found at {method_path}")
def _import_and_test(self, name, class_name, *args, **kwargs):
"""Test the import of the method and its functions."""
# Code structure
module = importlib.import_module(f"src.methods.{name}")
method = module.__getattribute__(class_name)(*args, **kwargs)
for fn in ["fit", "predict"]:
_ = method.__getattribute__(fn)
# Functions inputs and outputs
N, D = 10, 3
training_data = np.random.rand(N, D)
training_labels = np.random.randint(0, D, N)
test_data = np.random.rand(N, D)
with self.no_print():
pred_labels = method.fit(training_data, training_labels)
self.assertIsInstance(pred_labels, np.ndarray, f"{name}.{class_name}.fit() should output an array, not {type(pred_labels)}")
self.assertEqual(pred_labels.shape, training_labels.shape, f"{name}.{class_name}.fit() output has wrong shape ({pred_labels.shape} != {training_labels.shape})")
with self.no_print():
pred_labels = method.predict(test_data)
self.assertIsInstance(pred_labels, np.ndarray, f"{name}.{class_name}.predict() should output an array, not {type(pred_labels)}")
self.assertEqual(pred_labels.shape, training_labels.shape, f"{name}.{class_name}.predict() output has wrong shape ({pred_labels.shape} != {training_labels.shape})")
return method
def test_2_dummy_methods(self):
"""Test the dummy methods."""
self.title("Testing dummy methods")
_ = self._import_and_test("dummy_methods", "DummyClassifier",
arg1=1)
def test_4a_pca(self):
self.title("Testing PCA")
# Code structure
module = importlib.import_module("src.methods.pca")
pca = module.__getattribute__("PCA")(1)
for fn in ("find_principal_components", "reduce_dimension"):
_ = pca.__getattribute__(fn)
# Functions inputs and outputs
N, D, d = 10, 5, 2
pca = module.__getattribute__("PCA")(d)
data = np.random.rand(N, D)
with self.no_print():
exvar = pca.find_principal_components(data)
self.assertIsInstance(exvar, float, f"pca.PCA.find_principal_components() should output a float, not {type(exvar)}")
self.assertIsInstance(pca.mean, np.ndarray, f"pca.PCA.mean should be an array, not {type(pca.mean)}")
self.assertIn(pca.mean.shape, ((D,), (1,D)), f"pca.PCA.mean has wrong shape ({pca.mean.shape} != {(D,)})")
self.assertIsInstance(pca.W, np.ndarray, f"pca.PCA.W should be an array, not {type(pca.W)}")
self.assertEqual(pca.W.shape, (D, d), f"pca.PCA.W has wrong shape ({pca.W.shape} != {(D, d)})")
with self.no_print():
Y = pca.reduce_dimension(data)
self.assertIsInstance(Y, np.ndarray, f"pca.PCA.reduce_dimension() should output an array, not {type(Y)}")
self.assertEqual(Y.shape, (N, d), f"pca.PCA.reduce_dimension() output has wrong shape ({Y.shape} != {(N, d)})")
# Test on easy dummy data
N, D, d = 10, 2, 1
pca = module.__getattribute__("PCA")(d)
data = np.array([[2.77, 1.67], [1.96, 1.26], [ 0.67, 0.51], [0.99, 1.17], [-0.51, 0.21],
[0.12, 0.35], [ 2.46, 1.66], [2.05, 1.52], [1.51, 1.37], [ 2.09, 1.47]])
proj = np.array([-1.46, -0.55, 0.94, 0.35, 2.12, 1.50, -1.18, -0.75, -0.20, -0.76]).reshape(N, d)
with self.no_print():
exvar = pca.find_principal_components(data)
Y = pca.reduce_dimension(data)
self.assertGreater(exvar, 95., f"pca.PCA.find_principal_components() is not working on dummy data")
self.assertLess(np.linalg.norm(pca.mean - [1.41, 1.12]), 0.01, f"pca.PCA.find_principal_components() is not working on dummy data")
self.assertLess(np.linalg.norm(pca.W - [[-0.89], [-0.45]]), 0.01, f"pca.PCA.find_principal_components() is not working on dummy data")
self.assertLess(np.abs(Y - proj).max(), 0.01, f"pca.PCA.reduce_dimension() is not working on dummy data")
def test_4b_deep_network(self):
self.title("Testing deep-network")
# For dummy data
D, C = 10, 3
lr, epochs, bs = 0.01, 2, 8
# Code structure
trainer = self._import_and_test("deep_network", "Trainer", nn.Linear(3, 3), lr, epochs, bs)
for fn in ["train_all", "train_one_epoch", "predict_torch"]:
_ = trainer.__getattribute__(fn)
module = importlib.import_module("src.methods.deep_network")
for nn_type in ["MLP", "CNN"]:
model = module.__getattribute__(nn_type)(D if nn_type == "MLP" else 1, C)
trainer = module.__getattribute__("Trainer")(model, lr, epochs, bs)
# Functions inputs/outputs
N = 50
if nn_type == "MLP":
train_dataset = TensorDataset(torch.randn(N, D), torch.randint(0, C, (N,)))
test_dataset = TensorDataset(torch.randn(N, D))
elif nn_type == "CNN":
train_dataset = TensorDataset(torch.randn(N, 1, 32, 32), torch.randint(0, C, (N,)))
test_dataset = TensorDataset(torch.randn(N, 1, 32, 32))
train_dataloader = DataLoader(train_dataset, batch_size=bs, shuffle=True)
test_dataloader = DataLoader(test_dataset, batch_size=bs, shuffle=False)
# Test Network
with self.no_print():
x, _ = next(iter(train_dataloader))
preds = model(x)
self.assertIsInstance(preds, torch.Tensor, f"deep_network.{nn_type}.forward() should output a tensor, not {type(preds)}")
self.assertEqual(preds.shape, (bs, C), f"deep_network.{nn_type}.forward() output has wrong shape ({preds.shape} != {(bs, C)})")
# Test Trainer
with self.no_print():
trainer.train_all(train_dataloader)
pred_labels_test_torch = trainer.predict_torch(test_dataloader)
self.assertIsInstance(pred_labels_test_torch, torch.Tensor, f"deep_network.Trainer.predict_torch() should output a tensor, not {type(pred_labels_test_torch)}")
self.assertEqual(pred_labels_test_torch.shape, (N,), f"deep_network.Trainer.predict_torch() output has wrong shape ({pred_labels_test_torch.shape} != {(N,)})")
def warn(msg):
print(f"\33[33m/!\\ Warning: {msg}\33[39m")