def test_getdataloaders_dataloader_constructor_dataset_arg_is_correct_type(self):
def _dataloader_side_effect(*a, **kw):
dataset, = a
self.assertTrue(
isinstance(dataset, self.imagefolder_class)
)
self.mock_dataloader.side_effect = _dataloader_side_effect
train.getdataloaders('test_datadir', batch_size=1)
def test_imagefolder_is_instantiated_with_tranforms(self):
def _instantiate_side_effect(*a, **kw):
self.assertIn('transform', kw)
transform = kw['transform']
self.assertIn(
type(transform), [
torchvision.transforms.Compose
]
)
self.imagefolder_mock.side_effect = _instantiate_side_effect
train.getdataloaders(self.test_datadir, batch_size=self.test_batch_size)
def test_getdataloaders_dataloader_constructor_dataset_shuffle_arg(self):
shuffles = []
def _dataloader_side_effect(*a, **kw):
if 'shuffle' in kw and kw['shuffle']:
shuffles.append(True)
else:
shuffles.append(False)
self.mock_dataloader.side_effect = _dataloader_side_effect
train.getdataloaders('test_datadir', batch_size=1)
self.assertTrue(any(shuffles))
def test_imagefolder_is_instantiated_checking_path_arg(self):
observed_paths = set()
def _instantiate_side_effect(*a, **kw):
path, = a
observed_paths.add(path)
self.imagefolder_mock.side_effect = _instantiate_side_effect
train.getdataloaders(self.test_datadir, batch_size=self.test_batch_size)
expected_paths = set([
self.test_datadir + '/' + x for x in ('test', 'valid', 'train')
])
self.assertEqual(
expected_paths, observed_paths
)
def test_getdataloaders_dataloader_constructor_dataset_num_workers(self):
num_workers = []
def _dataloader_side_effect(*a, **kw):
self.assertIn('num_workers', kw)
num_workers.append(
kw['num_workers']
)
self.mock_dataloader.side_effect = _dataloader_side_effect
train.getdataloaders('test_datadir', batch_size=1)
self.assertEqual(
self.mock_dataloader.call_count,
3
)
self.assertTrue(
all([
x == train._DATALOADER_NUM_WORKERS for x in num_workers
])
)
def test_getdataloaders_dataloader_constructor_batch_size_arg(self):
batch_sizes = []
def _dataloader_side_effect(*a, **kw):
self.assertIn('batch_size', kw)
batch_sizes.append(
kw['batch_size']
)
self.mock_dataloader.side_effect = _dataloader_side_effect
test_batch_size = unittest.mock.sentinel.batch_size
train.getdataloaders('test_datadir', batch_size=test_batch_size)
self.assertEqual(
self.mock_dataloader.call_count,
3
)
self.assertTrue(
all([
x == test_batch_size for x in batch_sizes
])
)
def test_imagefolder_is_instantiated(self):
train.getdataloaders(self.test_datadir, batch_size=self.test_batch_size)
self.assertEqual(self.imagefolder_mock.call_count, 3)
def test_getdataloaders_retun_type(self):
train_ret, val_ret, test_ret, = train.getdataloaders('test_datadir', batch_size=1)
for r in [train_ret, val_ret, test_ret]:
self.assertTrue(
isinstance(r, self.dataloader_class)
)