def setUpClass(cls) -> None:
cls.ds = TestDatasets()
cls.meta_tensor_dataset = MetaDataset(cls.ds.tensor_dataset)
cls.meta_str_dataset = MetaDataset(cls.ds.str_dataset)
cls.meta_alpha_dataset = MetaDataset(cls.ds.alphabet_dataset)
cls.mnist_dataset = MetaDataset(cls.ds.get_mnist())
cls.omniglot_dataset = MetaDataset(cls.ds.get_omniglot())
def test_fails_with_non_torch_dataset(self):
try:
MetaDataset(np.random.randn(100, 100))
return False
except TypeError:
return True
finally:
return False
def test_load_data(self):
data = torch.randn(NUM_DATA, X_SHAPE)
labels = torch.randint(0, Y_SHAPE, (NUM_DATA, ))
dataset = MetaDataset(TensorDataset(data, labels))
task_dataset = TaskDataset(dataset,
task_transforms=[LoadData(dataset)],
num_tasks=NUM_TASKS)
for task in task_dataset:
self.assertTrue(isinstance(task[0], torch.Tensor))
self.assertTrue(isinstance(task[1], torch.Tensor))
def __init__(self, root=None, transforms=None):
img_folder = ImageFolder(root=root)
meta_data = MetaDataset(img_folder)
self.img_list = img_folder.imgs
self.labels_to_indices = meta_data.labels_to_indices
self.indices_to_labels = meta_data.indices_to_labels
self.labels = [0, 1, 2, 3, 4] # labels in dataset
self.transforms = transforms
def test_filter_labels(self):
data = torch.randn(NUM_DATA, X_SHAPE)
labels = torch.randint(0, Y_SHAPE, (NUM_DATA, ))
chosen_labels = random.sample(list(range(Y_SHAPE)), k=Y_SHAPE // 2)
dataset = MetaDataset(TensorDataset(data, labels))
task_dataset = TaskDataset(dataset,
task_transforms=[
FilterLabels(dataset, chosen_labels),
LoadData(dataset)
],
num_tasks=NUM_TASKS)
for task in task_dataset:
for label in task[1]:
self.assertTrue(label in chosen_labels)
def test_n_ways(self):
data = torch.randn(NUM_DATA, X_SHAPE)
labels = torch.randint(0, Y_SHAPE, (NUM_DATA, ))
dataset = MetaDataset(TensorDataset(data, labels))
for ways in range(1, 10):
task_dataset = TaskDataset(
dataset,
task_transforms=[NWays(dataset, n=ways),
LoadData(dataset)],
num_tasks=NUM_TASKS)
for task in task_dataset:
bins = task[1].bincount()
num_classes = len(bins) - (bins == 0).sum()
self.assertEqual(num_classes, ways)
def test_remap_labels(self):
data = torch.randn(NUM_DATA, X_SHAPE)
labels = torch.randint(0, Y_SHAPE, (NUM_DATA, ))
dataset = MetaDataset(TensorDataset(data, labels))
for ways in range(1, 5):
task_dataset = TaskDataset(dataset,
task_transforms=[
NWays(dataset, ways),
LoadData(dataset),
RemapLabels(dataset)
],
num_tasks=NUM_TASKS)
for task in task_dataset:
for label in range(ways):
self.assertTrue(label in task[1])
def __init__(self):
model = Net()
self.loss_fn = nn.CrossEntropyLoss()
model.to(device)
self.encoder = BertModel.from_pretrained('bert-base-chinese')
self.meta_model = l2l.algorithms.MAML(model, lr=1e-2, first_order=True)
self.optim = AdamW(self.meta_model.parameters(), lr=5e-3)
# text_train = l2l.text.datasets.NewsClassification(root=download_location, download=True)
# train_gen = l2l.text.datasets.TaskGenerator(text_train, ways=ways)
X, Y = load_data()
self.dataset = TensorDataset(T.LongTensor(X), T.LongTensor(Y))
self.metaset = MetaDataset(self.dataset)
self.task_generator = TaskGenerator(self.metaset,
ways=15,
shots=10,
classes=None,
tasks=1000)
def test_k_shots(self):
data = torch.randn(NUM_DATA, X_SHAPE)
labels = torch.randint(0, Y_SHAPE, (NUM_DATA, ))
dataset = MetaDataset(TensorDataset(data, labels))
for replacement in [False, True]:
for shots in range(1, 10):
task_dataset = TaskDataset(dataset,
task_transforms=[
KShots(dataset,
k=shots,
replacement=replacement),
LoadData(dataset)
],
num_tasks=NUM_TASKS)
for task in task_dataset:
bins = task[1].bincount()
correct = (bins == shots).sum()
self.assertEqual(correct, Y_SHAPE)
def setUpClass(cls) -> None:
cls.ds = TestDatasets()
cls.meta_tensor_dataset = MetaDataset(cls.ds.tensor_dataset)
cls.meta_str_dataset = MetaDataset(cls.ds.str_dataset)
cls.meta_alpha_dataset = MetaDataset(cls.ds.alphabet_dataset)