|
|
import numpy as np |
|
|
import pytest |
|
|
|
|
|
from .utils import FakeDataset, build_fake_dataset |
|
|
|
|
|
FLOWERS102_TESTSET_NUM = 5 * 102 |
|
|
|
|
|
|
|
|
def build_flowers102_fake_dataset(dataset_type: str, **kwargs) -> FakeDataset: |
|
|
dataset_cfg = dict(type=dataset_type.replace('Flowers102', 'FakeDataset'), |
|
|
x_shape=(3, 32, 32), |
|
|
y_range=(0, 101), |
|
|
nums=FLOWERS102_TESTSET_NUM, |
|
|
**kwargs) |
|
|
|
|
|
return build_fake_dataset(dataset_cfg) |
|
|
|
|
|
|
|
|
@pytest.mark.parametrize('dataset_type', ['Flowers102']) |
|
|
def test_xy(dataset_type: str) -> None: |
|
|
flowers102 = build_flowers102_fake_dataset(dataset_type) |
|
|
|
|
|
xy = flowers102.get_xy() |
|
|
x, y = xy |
|
|
assert len(x) == len(y) |
|
|
assert isinstance(y[0], int) |
|
|
|
|
|
old_x = x.copy() |
|
|
old_y = y.copy() |
|
|
|
|
|
flowers102.set_xy(xy) |
|
|
assert all( |
|
|
[np.array_equal(nx, ox) for nx, ox in zip(flowers102.data, old_x)]) |
|
|
assert flowers102.targets == old_y |
|
|
|
|
|
x = x[:flowers102.num_classes] |
|
|
y = y[:flowers102.num_classes] |
|
|
flowers102.set_xy((x, y)) |
|
|
assert all([np.array_equal(nx, ox) for nx, ox in zip(flowers102.data, x)]) |
|
|
assert flowers102.targets == y |
|
|
assert flowers102.num_classes == len(set(y)) |
|
|
assert len(flowers102.data.shape) == 4 |
|
|
|
|
|
|
|
|
@pytest.mark.parametrize('dataset_type', ['PoisonLabelFlowers102']) |
|
|
@pytest.mark.parametrize('poison_label', [-1, 5, 102]) |
|
|
def test_poison_label(poison_label: int, dataset_type: str) -> None: |
|
|
kwargs = dict(poison_label=poison_label) |
|
|
|
|
|
if poison_label < 0 or poison_label >= 43: |
|
|
with pytest.raises(ValueError): |
|
|
_ = build_flowers102_fake_dataset(dataset_type, **kwargs) |
|
|
return |
|
|
flowers102 = build_flowers102_fake_dataset(dataset_type, **kwargs) |
|
|
assert flowers102.poison_label == poison_label |
|
|
|
|
|
assert flowers102.num_classes == 1 |
|
|
assert all(map(lambda x: x == poison_label, flowers102.targets)) |
|
|
assert len(flowers102.data.shape) == 4 |
|
|
|
|
|
|
|
|
@pytest.mark.parametrize('dataset_type', ['RatioPoisonLabelFlowers102']) |
|
|
@pytest.mark.parametrize('poison_label', [-1, 5, 102]) |
|
|
@pytest.mark.parametrize('ratio', [0, 0.2, 1, 1.2]) |
|
|
def test_ratio_poison_label(ratio: float, poison_label: int, |
|
|
dataset_type: str) -> None: |
|
|
kwargs = dict(ratio=ratio, poison_label=poison_label) |
|
|
|
|
|
if (poison_label < 0 or poison_label >= 102) or \ |
|
|
(ratio <= 0 or ratio > 1): |
|
|
with pytest.raises(ValueError): |
|
|
_ = build_flowers102_fake_dataset(dataset_type, **kwargs) |
|
|
return |
|
|
flowers102 = build_flowers102_fake_dataset(dataset_type, **kwargs) |
|
|
assert flowers102.poison_label == poison_label |
|
|
|
|
|
assert len(flowers102) == round(FLOWERS102_TESTSET_NUM * ratio) |
|
|
assert flowers102.num_classes == 1 |
|
|
assert all(map(lambda x: x == poison_label, flowers102.targets)) |
|
|
assert len(flowers102.data.shape) == 4 |
|
|
|
