def test_train_iteration():
""" Tests return types and sizes of train_iteration """
model = test_model()
image_dir = 'tests/test_data/images_10x'
data = prep.prep_data(pd.read_csv('tests/test_data/10x_labels_4.csv'),
image_dir)
image_dir = 'tests/test_data/images_10x'
opt = torch.optim.Adam(model.parameters(), lr=.01)
crit = torch.nn.CrossEntropyLoss()
device = torch.device('cpu')
transforms = torchvision.transforms.Compose([
torchvision.transforms.ToPILImage(),
torchvision.transforms.RandomRotation((-180, 180)),
torchvision.transforms.CenterCrop((325)),
torchvision.transforms.ToTensor()
])
dataset = prep.tenX_dataset(data, image_dir, transforms)
iterator = torch.utils.data.DataLoader(dataset, shuffle=True, batch_size=1)
loss, acc, pred, label = construct.train_iteration(model, iterator, opt,
crit, device)
assert isinstance(loss, float), f'train_iteration error,\
wrong loss dtype {loss.type()}'
assert isinstance(acc, float),\
f'train_iteration error, wrong acc dtype {acc.type()}'
assert pred.size() == torch.Size([1, 2]),\
f'train_iteration error, prediction wrong shape/size {pred.size()}'
assert label.size() == torch.Size([1, 2]),\
f'train_iteration error, label wrong shape/size {label.size()}'
def test_get_predictions():
"""
Tests return types and sizes of get_predictions
"""
model = test_model()
image_dir = 'tests/test_data/images_10x'
data = prep.prep_data(pd.read_csv('tests/test_data/10x_labels_4.csv'),
image_dir)
transforms = torchvision.transforms.Compose([
torchvision.transforms.ToPILImage(),
torchvision.transforms.RandomRotation((-180, 180)),
torchvision.transforms.CenterCrop((325)),
torchvision.transforms.ToTensor()
])
dataset = prep.tenX_dataset(data, image_dir, transforms)
images, labels, predictions, weights, acc =\
construct.get_predictions(1, model, dataset)
assert labels.size() == torch.Size([12, 2]),\
f'{"get_predictions error, incorrect labels dimensions"}'
assert isinstance(acc.item(), float),\
f'{"get_predictions error, incorrect accuracy dtype"}'
assert weights.size() == torch.Size([12, 2]),\
f'{"get_predictions error, incorrect weight dimensions"}'
assert predictions.size() == torch.Size([12]),\
f'{"get_predictions error, incorrect predictions dimensions"}'
assert images.size() == torch.Size([12, 3, 325, 325]),\
f'{"get_predictions error, incorrect image"}'
def test_tenX_dataset():
"""Test if the class tenX_dataset can generate correct output."""
# Load the inputs for tenX_dataset.
image_dir = 'tests/test_data/images_10x'
labels = prep.prep_data(pd.read_csv('tests/test_data/10x_labels_5.csv'),
image_dir)
# To make the test more simple, define the "transform"
# as a function that returns
# the input directly without doing anything.
def transforms(image):
return image
# Create an object tenX.
tenX = prep.tenX_dataset(labels, image_dir, transforms)
# check if the class tenX_dataset can generate the correct
# output by comparing the actual output with the expected output.
assert len(tenX) == 10, "The len method in class tenX_dataset is broken!"
assert tenX[1]['image'].size == 1082880,\
"The getitem method in class tenX_dataset is broken!"
assert np.array_equal(tenX[1]['shape'], np.array([0, 0, 0, 1])),\
"The getitem method in class tenX_dataset is broken!"
assert np.array_equal(tenX[1]['color'], np.array([0, 0, 1, 0])),\
"The getitem method in class tenX_dataset is broken!"
assert tenX[1]['plastic'][0] == 1,\
"The getitem method in class tenX_dataset is broken!"
return None
def test_prep_data_1():
"""
Test if the prep_data function can generate correct output.
"""
# Load the csv file as the input data frame.
input_df = pd.read_csv('tests/test_data/10x_labels_5.csv')
image_dir = 'tests/test_data/images_10x'
# Call the prep_data function and get the actual output "result".
result = prep.prep_data(input_df, image_dir)
# Load the output data frame from a csv file.
output_df = pd.read_csv('tests/test_data/prep_data_output.csv')
# Modify the format of the output data
# frame to make it the expected output.
for i, rowi in output_df['Sample'].iteritems():
output_df['Sample'].loc[i] = rowi.split(',')
for j, rowj in output_df['Color'].iteritems():
output_df['Color'].loc[j] = np.fromstring(rowj, dtype=int, sep=' ')
for k, rowk in output_df['Shape'].iteritems():
output_df['Shape'].loc[k] = np.fromstring(rowk, dtype=int, sep=' ')
output_df['isPlastic'] = [
np.array([1, 0]),
np.array([1, 0]),
np.array([0, 1]),
np.array([1, 0]),
np.array([0, 1]),
np.array([1, 0]),
np.array([1, 0]),
np.array([1, 0]),
np.array([0, 1]),
np.array([0, 1])
]
output_df['Color'] = [
np.array([0, 0, 1, 0]),
np.array([0, 0, 1, 0]),
np.array([0, 0, 0, 1]),
np.array([1, 0, 0, 0]),
np.array([0, 1, 0, 0]),
np.array([0, 1, 0, 0]),
np.array([0, 0, 1, 0]),
np.array([0, 0, 1, 0]),
np.array([0, 1, 0, 0]),
np.array([0, 1, 0, 0])
]
output_df['Shape'] = [
np.array([1, 0, 0, 0]),
np.array([0, 0, 0, 1]),
np.array([0, 0, 0, 1]),
np.array([0, 1, 0, 0]),
np.array([0, 1, 0, 0]),
np.array([1, 0, 0, 0]),
np.array([0, 0, 0, 1]),
np.array([0, 0, 1, 0]),
np.array([0, 0, 0, 1]),
np.array([1, 0, 0, 0])
]
# Check if the expected output data frame is
# the same as the actual output data frame.
result = result.drop(columns=['index'])
assert output_df.equals(result), "The prep_data function is broken!"
return None
def test_kfold():
model = test_model()
n_splits = 5
epochs = 2
batch_size = 1
transforms = torchvision.transforms.Compose([
torchvision.transforms.ToPILImage(),
torchvision.transforms.RandomRotation((-180, 180)),
torchvision.transforms.CenterCrop((325)),
torchvision.transforms.ToTensor()
])
crit = torch.nn.CrossEntropyLoss()
image_dir = 'tests/test_data/images_10x'
df = prep.prep_data(pd.read_csv('tests/test_data/10x_labels_4.csv'),
image_dir)
data = prep.tenX_dataset(df, image_dir, transforms)
device = torch.device('cpu')
models, losses, train_accs,\
naive_accs, test_accs = construct.k_fold(n_splits,
epochs, batch_size,
transforms, crit,
model,
data, device)
assert len(losses) is n_splits, f'kfold failed,\
expected {n_splits} loss values, but got {len(losses)}'
assert len(train_accs) is n_splits, f'kfold failed,\
expected {n_splits} train lists, but got {len(train_accs)}'
assert len(train_accs[0]) is (epochs + 1), f'kflod failed,\
expected {epochs + 1} length train lists, but got {len(train_accs[0])}'
assert len(naive_accs) is n_splits, f'kfold failed,\
expected {n_splits} naive accs, but got {len(naive_accs)}'
assert len(test_accs) is n_splits, f'kfold failed,\
expected {n_splits} val accs, but got {len(test_accs)}'
assert isinstance(test_accs[0].item(), float), f'kflod failed,\
expected float, but got {test_accs[0].item().type()}'
assert isinstance(naive_accs[0].item(), float), f'kflod failed,\
def test_train():
"""
Tests returns types and sizes of train
"""
image_dir = 'tests/test_data/images_10x'
data = prep.prep_data(pd.read_csv('tests/test_data/10x_labels_4.csv'),
image_dir)
transforms = torchvision.transforms.Compose([
torchvision.transforms.ToPILImage(),
torchvision.transforms.RandomRotation((-180, 180)),
torchvision.transforms.CenterCrop((325)),
torchvision.transforms.ToTensor()
])
dataset = prep.tenX_dataset(data, image_dir, transforms)
mod, loss, acc = construct.train(2, 1, dataset)
assert isinstance(loss[0], float), f'train failed,\
returned wrong dataype for loss: {loss.type()}'
assert isinstance(acc[0], float), f'train failed,\
returned wrong dataype for accuracy: {acc.type()}'
assert len(loss) == 3, f'train failed,\
def test_prep_data_2():
"""
Test if the prep_data function is responsive
to a wrong datatype of the input.
"""
image_dir = 'tests/test_data/images_10x'
input_1 = 10
test1 = False
try:
prep.prep_data(input_1, image_dir)
except Exception as e:
assert isinstance(e, TypeError), "Wrong type of error."
test1 = True
assert test1, "Test failed!\
The prep_data function is not\
responsive to the wrong input datatype 'int'."
input_2 = 1.2
test2 = False
try:
prep.prep_data(input_2, image_dir)
except Exception as e:
assert isinstance(e, TypeError), "Wrong type of error."
test2 = True
assert test2, "Test failed!\
The prep_data function is not\
responsive to the wrong input datatype 'float'."
input_3 = [1, 2, 3]
test3 = False
try:
prep.prep_data(input_3, image_dir)
except Exception as e:
assert isinstance(e, TypeError), "Wrong type of error."
test3 = True
assert test3, "Test failed!\
The prep_data function is not\
responsive to the wrong input datatype 'list'."
input_4 = (1, 2, 3)
test4 = False
try:
prep.prep_data(input_4, image_dir)
except Exception as e:
assert isinstance(e, TypeError), "Wrong type of error."
test4 = True
assert test4, "Test failed!\
The prep_data function is not\
responsive to the wrong input datatype 'tuple'."
input_5 = 'input'
test5 = False
try:
prep.prep_data(input_5, image_dir)
except Exception as e:
assert isinstance(e, TypeError), "Wrong type of error."
test5 = True
assert test5, "Test failed!\
The prep_data function is not\
responsive to the wrong input datatype 'str'."
input_6 = False
test6 = False
try:
prep.prep_data(input_6, image_dir)
except Exception as e:
assert isinstance(e, TypeError), "Wrong type of error."
test6 = True
assert test6, "Test failed!\
The prep_data function is not\
responsive to the wrong input datatype 'bool'."
return None