Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

[dask] Rework base margin test. #6627

Merged
merged 1 commit into from
Jan 22, 2021
Merged

[dask] Rework base margin test. #6627

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Jump to
Jump to file
Failed to load files.
Diff view
Diff view
84 changes: 23 additions & 61 deletions tests/python/test_with_dask.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -149,68 +149,30 @@ def test_dask_predict_shape_infer() -> None:


@pytest.mark.parametrize("tree_method", ["hist", "approx"])
def test_boost_from_prediction(tree_method: str) -> None:
if tree_method == 'approx':
pytest.xfail(reason='test_boost_from_prediction[approx] is flaky')

def test_boost_from_prediction(tree_method: str, client: "Client") -> None:
from sklearn.datasets import load_breast_cancer
X, y = load_breast_cancer(return_X_y=True)

X_ = dd.from_array(X, chunksize=100)
y_ = dd.from_array(y, chunksize=100)

with LocalCluster(n_workers=4) as cluster:
with Client(cluster) as _:
model_0 = xgb.dask.DaskXGBClassifier(
learning_rate=0.3,
random_state=123,
n_estimators=4,
tree_method=tree_method,
)
model_0.fit(X=X_, y=y_)
margin = model_0.predict(X_, output_margin=True)

model_1 = xgb.dask.DaskXGBClassifier(
learning_rate=0.3,
random_state=123,
n_estimators=4,
tree_method=tree_method,
)
model_1.fit(X=X_, y=y_, base_margin=margin)
predictions_1 = model_1.predict(X_, base_margin=margin)
proba_1 = model_1.predict_proba(X_, base_margin=margin)

cls_2 = xgb.dask.DaskXGBClassifier(
learning_rate=0.3,
random_state=123,
n_estimators=8,
tree_method=tree_method,
)
cls_2.fit(X=X_, y=y_)
predictions_2 = cls_2.predict(X_)
proba_2 = cls_2.predict_proba(X_)

cls_3 = xgb.dask.DaskXGBClassifier(
learning_rate=0.3,
random_state=123,
n_estimators=8,
tree_method=tree_method,
)
cls_3.fit(X=X_, y=y_)
proba_3 = cls_3.predict_proba(X_)

# compute variance of probability percentages between two of the
# same model, use this to check to make sure approx is functioning
# within normal parameters
expected_variance = np.max(np.abs(proba_3 - proba_2)).compute()

if expected_variance > 0:
margin_variance = np.max(np.abs(proba_1 - proba_2)).compute()
# Ensure the margin variance is less than the expected variance + 10%
assert np.all(margin_variance <= expected_variance + .1)
else:
np.testing.assert_equal(predictions_1.compute(), predictions_2.compute())
np.testing.assert_almost_equal(proba_1.compute(), proba_2.compute())
X_, y_ = load_breast_cancer(return_X_y=True)

X, y = dd.from_array(X_, chunksize=100), dd.from_array(y_, chunksize=100)
model_0 = xgb.dask.DaskXGBClassifier(
learning_rate=0.3, random_state=0, n_estimators=4,
tree_method=tree_method)
model_0.fit(X=X, y=y)
margin = model_0.predict(X, output_margin=True)

model_1 = xgb.dask.DaskXGBClassifier(
learning_rate=0.3, random_state=0, n_estimators=4,
tree_method=tree_method)
model_1.fit(X=X, y=y, base_margin=margin)
predictions_1 = model_1.predict(X, base_margin=margin)

cls_2 = xgb.dask.DaskXGBClassifier(
learning_rate=0.3, random_state=0, n_estimators=8,
tree_method=tree_method)
cls_2.fit(X=X, y=y)
predictions_2 = cls_2.predict(X)

assert np.all(predictions_1.compute() == predictions_2.compute())


def test_dask_missing_value_reg() -> None:
Expand Down