def test_multilabel_accuracy_mean(
outputs_list: Iterable[torch.Tensor],
targets_list: Iterable[torch.Tensor],
thresholds: Union[float, torch.Tensor],
true_values_list: Iterable[float],
) -> None:
"""
This test checks that all the intermediate metrics values are correct during accumulation.
Args:
outputs_list: list of output tensors
targets_list: list of true answer tensors
thresholds: threshold
true_values_list: true intermediate metric results
"""
metric = MultilabelAccuracyMetric(threshold=thresholds)
for outputs, targets, true_value in zip(outputs_list, targets_list,
true_values_list):
metric.update(outputs=outputs, targets=targets)
mean, _ = metric.compute()
assert np.isclose(mean, true_value)
def test_multilabel_accuracy_std(
outputs_list: Iterable[torch.Tensor],
targets_list: Iterable[torch.Tensor],
thresholds: Union[float, torch.Tensor],
true_values_list: Iterable[float],
) -> None:
"""
This test checks that all the intermediate metrics values are correct during accumulation.
Note that now `accuracy/std` is not std exactly so it can fail if you fix it.
Args:
outputs_list: list of output tensors
targets_list: list of true answer tensors
thresholds: threshold
true_values_list: true intermediate metric results
"""
metric = MultilabelAccuracyMetric(threshold=thresholds)
for outputs, targets, true_value in zip(outputs_list, targets_list,
true_values_list):
metric.update(outputs=outputs, targets=targets)
_, std = metric.compute()
assert np.isclose(std, true_value)
def test_multilabel_accuracy(
outputs: torch.Tensor,
targets: torch.Tensor,
thresholds: Union[float, torch.Tensor],
true_values: Dict[str, float],
) -> None:
"""
Test multilabel accuracy metric with single and multiple thresholds.
Note that now `accuracy/std` is not std exactly so it can fail if you fix it.
Args:
outputs: tensor of outputs
targets: tensor of true answers
thresholds: thresholds for multilabel classification
true_values: expected metric value
"""
metric = MultilabelAccuracyMetric(threshold=thresholds)
metric.update(outputs=outputs, targets=targets)
values = metric.compute_key_value()
for key in true_values.keys():
assert key in true_values
assert np.isclose(true_values[key], values[key])
def __init__(
self,
input_key: str,
target_key: str,
threshold: Union[float, torch.Tensor] = 0.5,
log_on_batch: bool = True,
prefix: str = None,
suffix: str = None,
):
"""Init."""
super().__init__(
metric=MultilabelAccuracyMetric(threshold=threshold, prefix=prefix, suffix=suffix),
input_key=input_key,
target_key=target_key,
log_on_batch=log_on_batch,
)