def test_get_dataset(data_path):
"""Test splitting with default size of datasets."""
train, test, valid = get_datasets(
data_path=data_path,
nb_nodes=185,
task_type="classification",
nb_classes=2,
split=None,
k_fold=None,
seed=1234,
)
assert 164 == len(train)
assert 24 == len(test)
assert 48 == len(valid)
def test_kfold(data_path):
"""Test kfold splitting."""
train, test, valid = get_datasets(
data_path=data_path,
nb_nodes=185,
task_type="classification",
nb_classes=2,
split=None,
k_fold=[7, 1, 2],
seed=1234,
)
assert 164 == len(train)
assert 24 == len(test)
assert 48 == len(valid)
def batch(data_path):
"""Define a batch for the KrasHras dataset."""
train, _, _ = get_datasets(
data_path=data_path,
nb_nodes=7,
task_type="classification",
nb_classes=2,
split=None,
k_fold=None,
seed=1234,
)
for batch in torch.utils.data.DataLoader(
train, shuffle=False, batch_size=25, drop_last=False
):
return batch
def test_indexing(data_path):
"""Test random access the generated graph dataset."""
train, _, _ = get_datasets(
data_path=data_path,
nb_nodes=185,
task_type="classification",
nb_classes=2,
split=None,
k_fold=None,
seed=1234,
)
prot = train[0]
prot_dims = [len(tr) for tr in prot]
# v, c, m, y
assert prot_dims == [185, 185, 185, 2]
def test_gaussian_augmentation(data_path):
"""Test splitting with default size of datasets."""
train, valid, test = get_datasets(
data_path=data_path,
nb_nodes=185,
task_type="classification",
nb_classes=2,
split=None,
augment=2,
k_fold=None,
seed=1234,
)
assert 164 * 2 == len(train)
assert 48 * 2 == len(test)
assert 24 == len(valid)
def test_validation(data_path, nn_kras):
"""Fit one epoch of train + test."""
train, test, valid = get_datasets(
data_path=data_path,
nb_nodes=7,
task_type="classification",
nb_classes=2,
split=None,
k_fold=None,
seed=1234,
)
validator = Validation(nn_kras, valid)
validator.validate()
stats = validator.compute_stats()
set(stats.keys()) == {
"recall",
"precision",
"accuracy",
"f_score",
}
def test_dataloading_batch(data_path):
"""Test transformation of input."""
train, _, _ = get_datasets(
data_path=data_path,
nb_nodes=185,
task_type="classification",
nb_classes=2,
split=None,
k_fold=None,
seed=1234,
)
trainloader = torch.utils.data.DataLoader(
train, shuffle=False, batch_size=25, drop_last=False
)
for batch in trainloader:
batch_dims = [len(tr) for tr in batch]
break
v = batch[0]
assert batch_dims == [25, 25, 25, 2]
assert v.shape == torch.Size([25, 185, 29])
def test_fit_epoch(data_path, nn_kras):
"""Fit one epoch of train + test."""
train, test, _ = get_datasets(
data_path=data_path,
nb_nodes=7,
task_type="classification",
nb_classes=2,
split=None,
k_fold=None,
seed=1234,
)
optimizer = torch.optim.Adam(nn_kras.parameters())
criterion = torch.nn.CrossEntropyLoss()
fit_network(
nn_kras,
train,
test,
optimizer,
criterion,
20,
epochs=1,
plot_every=2000,
)