def test_meta_dataloader():
dataset = Sinusoid(10, num_tasks=1000, noise_std=None)
meta_dataloader = MetaDataLoader(dataset, batch_size=4)
assert isinstance(meta_dataloader, DataLoader)
assert len(meta_dataloader) == 250 # 1000 / 4
batch = next(iter(meta_dataloader))
assert isinstance(batch, list)
assert len(batch) == 4
task = batch[0]
assert isinstance(task, Task)
assert len(task) == 10
def test_meta_dataloader_task_loader():
dataset = Sinusoid(10, num_tasks=1000, noise_std=None)
meta_dataloader = MetaDataLoader(dataset, batch_size=4)
batch = next(iter(meta_dataloader))
dataloader = DataLoader(batch[0], batch_size=5)
inputs, targets = next(iter(dataloader))
assert len(dataloader) == 2 # 10 / 5
# PyTorch dataloaders convert numpy array to tensors
assert isinstance(inputs, torch.Tensor)
assert isinstance(targets, torch.Tensor)
assert inputs.shape == (5, 1)
assert targets.shape == (5, 1)
def get_episode_loader(dataset,
datapath,
ways,
shots,
test_shots,
batch_size,
split,
download=True,
shuffle=True,
num_workers=0):
"""Create an episode data loader for a torchmeta dataset. Can also
include unlabelled data for semi-supervised learning.
dataset: String. Name of the dataset to use.
datapath: String. Path, where dataset are stored.
ways: Integer. Number of ways N.
shots: Integer. Number of shots K for support set.
test_shots: Integer. Number of images in query set.
batch_size: Integer. Number of tasks per iteration.
split: String. One of ['train', 'val', 'test']
download: Boolean. Whether to download the data.
shuffle: Boolean. Whether to shuffle episodes.
"""
# Select dataset
if dataset == 'omniglot':
dataset_func = omniglot
elif dataset == 'miniimagenet':
dataset_func = miniimagenet
elif dataset == 'tieredimagenet':
dataset_func = tieredimagenet
elif dataset == 'cub':
dataset_func = cub
elif dataset == 'cifar_fs':
dataset_func = cifar_fs
elif dataset == 'doublemnist':
dataset_func = doublemnist
elif dataset == 'triplemnist':
dataset_func = triplemnist
else:
raise ValueError("No such dataset available. Please choose from\
['omniglot', 'miniimagenet', 'tieredimagenet',\
'cub, cifar_fs, doublemnist, triplemnist']")
# Collect arguments that are the same for all possible sub-datasets
kwargs = {
'download': download,
'meta_train': split == 'train',
'meta_val': split == 'val',
'meta_test': split == 'test',
'shuffle': shuffle
}
# Create dataset for labelled images
dataset_name = dataset
dataset = dataset_func(datapath,
ways=ways,
shots=shots,
test_shots=test_shots,
**kwargs)
print('Supervised data loader for {}:{}.'.format(dataset_name, split))
# Standard supervised meta-learning dataloader
collate_fn = collate_task_batch if batch_size else collate_task
return MetaDataLoader(dataset,
batch_size=batch_size,
collate_fn=collate_fn,
num_workers=num_workers,
pin_memory=torch.cuda.is_available())