def test_wrong_input_shapes():
m = MedianAbsoluteError()
with pytest.raises(ValueError):
m.update((torch.rand(4, 1, 2), torch.rand(4, 1)))
with pytest.raises(ValueError):
m.update((torch.rand(4, 1), torch.rand(4, 1, 2)))
with pytest.raises(ValueError):
m.update(
(
torch.rand(4, 1, 2),
torch.rand(
4,
),
)
)
with pytest.raises(ValueError):
m.update(
(
torch.rand(
4,
),
torch.rand(4, 1, 2),
)
)
def test_integration_median_absolute_error_with_output_transform():
np.random.seed(1)
size = 105
np_y_pred = np.random.rand(size, 1)
np_y = np.random.rand(size, 1)
np.random.shuffle(np_y)
np_median_absolute_error = np.median(np.abs(np_y - np_y_pred))
batch_size = 15
def update_fn(engine, batch):
idx = (engine.state.iteration - 1) * batch_size
y_true_batch = np_y[idx : idx + batch_size]
y_pred_batch = np_y_pred[idx : idx + batch_size]
return idx, torch.from_numpy(y_pred_batch), torch.from_numpy(y_true_batch)
engine = Engine(update_fn)
m = MedianAbsoluteError(output_transform=lambda x: (x[1], x[2]))
m.attach(engine, "median_absolute_error")
data = list(range(size // batch_size))
median_absolute_error = engine.run(data, max_epochs=1).metrics[
"median_absolute_error"
]
assert np_median_absolute_error == pytest.approx(median_absolute_error)
def _test(n_epochs, metric_device):
metric_device = torch.device(metric_device)
n_iters = 80
size = 105
y_true = torch.rand(size=(size, )).to(device)
y_preds = torch.rand(size=(size, )).to(device)
def update(engine, i):
return (
y_preds[i * size:(i + 1) * size],
y_true[i * size:(i + 1) * size],
)
engine = Engine(update)
m = MedianAbsoluteError(device=metric_device)
m.attach(engine, "mae")
data = list(range(n_iters))
engine.run(data=data, max_epochs=n_epochs)
assert "mae" in engine.state.metrics
res = engine.state.metrics["mae"]
np_y_true = y_true.cpu().numpy().ravel()
np_y_preds = y_preds.cpu().numpy().ravel()
e = np.abs(np_y_true - np_y_preds)
np_res = np.median(e)
assert pytest.approx(res) == np_res
def _test(metric_device):
metric_device = torch.device(metric_device)
m = MedianAbsoluteError(device=metric_device)
torch.manual_seed(10 + rank)
size = 105
y_pred = torch.randint(1,
10,
size=(size, 1),
dtype=torch.double,
device=device)
y = torch.randint(1,
10,
size=(size, 1),
dtype=torch.double,
device=device)
m.update((y_pred, y))
# gather y_pred, y
y_pred = idist.all_gather(y_pred)
y = idist.all_gather(y)
np_y_pred = y_pred.cpu().numpy().ravel()
np_y = y.cpu().numpy().ravel()
res = m.compute()
np_median_absolute_error = np.median(np.abs(np_y - np_y_pred))
assert np_median_absolute_error == pytest.approx(res)
def test_zero_sample():
m = MedianAbsoluteError()
with pytest.raises(
NotComputableError,
match=
r"EpochMetric must have at least one example before it can be computed"
):
m.compute()
def test_wrong_input_shapes():
m = MedianAbsoluteError()
with pytest.raises(ValueError, match=r"Predictions should be of shape"):
m.update((torch.rand(4, 1, 2), torch.rand(4, 1)))
with pytest.raises(ValueError, match=r"Targets should be of shape"):
m.update((torch.rand(4, 1), torch.rand(4, 1, 2)))
with pytest.raises(ValueError, match=r"Predictions should be of shape"):
m.update((torch.rand(4, 1, 2), torch.rand(4)))
with pytest.raises(ValueError, match=r"Targets should be of shape"):
m.update((torch.rand(4), torch.rand(4, 1, 2)))
def _test(metric_device):
metric_device = torch.device(metric_device)
m = MedianAbsoluteError(device=metric_device)
torch.manual_seed(10 + rank)
size = 105
y_pred = torch.randint(1,
10,
size=(size, 1),
dtype=torch.double,
device=device)
y = torch.randint(1,
10,
size=(size, 1),
dtype=torch.double,
device=device)
m.update((y_pred, y))
# gather y_pred, y
y_pred = idist.all_gather(y_pred)
y = idist.all_gather(y)
np_y_pred = y_pred.cpu().numpy().ravel()
np_y = y.cpu().numpy().ravel()
res = m.compute()
e = np.abs(np_y - np_y_pred)
# The results between numpy.median() and torch.median() are Inconsistant
# when the length of the array/tensor is even. So this is a hack to avoid that.
# issue: https://github.com/pytorch/pytorch/issues/1837
if np_y_pred.shape[0] % 2 == 0:
e_prepend = np.insert(e, 0, e[0], axis=0)
np_res_prepend = np.median(e_prepend)
assert pytest.approx(res) == np_res_prepend
else:
np_res = np.median(e)
assert pytest.approx(res) == np_res
def test_median_absolute_error():
# See https://github.com/torch/torch7/pull/182
# For even number of elements, PyTorch returns middle element
# NumPy returns average of middle elements
# Size of dataset will be odd for these tests
size = 51
np_y_pred = np.random.rand(size)
np_y = np.random.rand(size)
np_median_absolute_error = np.median(np.abs(np_y - np_y_pred))
m = MedianAbsoluteError()
y_pred = torch.from_numpy(np_y_pred)
y = torch.from_numpy(np_y)
m.reset()
m.update((y_pred, y))
assert np_median_absolute_error == pytest.approx(m.compute())
def test_median_absolute_error_2():
np.random.seed(1)
size = 105
np_y_pred = np.random.rand(size, 1)
np_y = np.random.rand(size, 1)
np.random.shuffle(np_y)
np_median_absolute_error = np.median(np.abs(np_y - np_y_pred))
m = MedianAbsoluteError()
y_pred = torch.from_numpy(np_y_pred)
y = torch.from_numpy(np_y)
m.reset()
batch_size = 16
n_iters = size // batch_size + 1
for i in range(n_iters):
idx = i * batch_size
m.update((y_pred[idx : idx + batch_size], y[idx : idx + batch_size]))
assert np_median_absolute_error == pytest.approx(m.compute())
