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

[REVIEW] Update failing MNMG tests #3348

Merged
merged 8 commits into from
Feb 8, 2021
Merged

[REVIEW] Update failing MNMG tests #3348

Show file tree
Hide file tree
Changes from 4 commits
Commits
Show all changes
8 commits
Select commit Hold shift + click to select a range
3f6e827
Update MNMG RF testing
viclafargue Jan 6, 2021
906fa8d
Update MNMG KNN testing
viclafargue Jan 6, 2021
d10bbb4
Merge branch 'branch-0.18' into fix-mnmg-failing-tests
viclafargue Jan 14, 2021
6164283
Update
viclafargue Jan 14, 2021
c675aa8
Lowering number of samples for RF test
viclafargue Jan 19, 2021
163eb18
Merge branch 'branch-0.18' into fix-mnmg-failing-tests
viclafargue Jan 27, 2021
ce3f70b
Merge branch 'branch-0.18' into fix-mnmg-failing-tests
viclafargue Jan 29, 2021
bf68814
Check style
viclafargue Jan 29, 2021
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion python/cuml/test/dask/test_nearest_neighbors.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -248,7 +248,7 @@ def test_one_query_partition(client):

X_train, _ = make_blobs(n_samples=4000,
n_features=16,
n_parts=4)
n_parts=8)

X_test, _ = make_blobs(n_samples=200,
n_features=16,
Expand Down
60 changes: 43 additions & 17 deletions python/cuml/test/dask/test_random_forest.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -75,18 +75,19 @@ def test_rf_classification_multi_class(partitions_per_worker, cluster):

# Use CUDA_VISIBLE_DEVICES to control the number of workers
c = Client(cluster)
n_workers = len(c.scheduler_info()['workers'])

try:

X, y = make_classification(n_samples=10000, n_features=20,
X, y = make_classification(n_samples=n_workers * 5000, n_features=20,
n_clusters_per_class=1, n_informative=10,
random_state=123, n_classes=15)

X = X.astype(np.float32)
y = y.astype(np.int32)

X_train, X_test, y_train, y_test = \
train_test_split(X, y, test_size=1000, random_state=123)
train_test_split(X, y, test_size=n_workers * 300, random_state=123)

cu_rf_params = {
'n_estimators': 25,
Expand All @@ -98,7 +99,7 @@ def test_rf_classification_multi_class(partitions_per_worker, cluster):
X_train_df, y_train_df = _prep_training_data(c, X_train, y_train,
partitions_per_worker)

cuml_mod = cuRFC_mg(**cu_rf_params)
cuml_mod = cuRFC_mg(**cu_rf_params, ignore_empty_partitions=True)
cuml_mod.fit(X_train_df, y_train_df)
X_test_dask_array = from_array(X_test)
cuml_preds_gpu = cuml_mod.predict(X_test_dask_array,
Expand All @@ -111,7 +112,7 @@ def test_rf_classification_multi_class(partitions_per_worker, cluster):
# Refer to issue : https://github.com/rapidsai/cuml/issues/2806 for
# more information on the threshold value.

assert acc_score_gpu >= 0.60
assert acc_score_gpu >= 0.55

finally:
c.close()
Expand All @@ -121,16 +122,18 @@ def test_rf_classification_multi_class(partitions_per_worker, cluster):
@pytest.mark.parametrize('partitions_per_worker', [5])
def test_rf_regression_dask_fil(partitions_per_worker,
dtype, client):
n_workers = len(client.scheduler_info()['workers'])

# Use CUDA_VISIBLE_DEVICES to control the number of workers
X, y = make_regression(n_samples=10000, n_features=20,
X, y = make_regression(n_samples=n_workers * 4000, n_features=20,
n_informative=10, random_state=123)

X = X.astype(dtype)
y = y.astype(dtype)

X_train, X_test, y_train, y_test = train_test_split(X, y,
test_size=1000,
random_state=123)
X_train, X_test, y_train, y_test = \
train_test_split(X, y, test_size=n_workers * 100,
random_state=123)

if dtype == np.float64:
pytest.xfail(reason=" Dask RF does not support np.float64 data")
Expand All @@ -155,7 +158,7 @@ def test_rf_regression_dask_fil(partitions_per_worker,
X_test_df = \
dask_cudf.from_cudf(X_cudf_test, npartitions=n_partitions)

cuml_mod = cuRFR_mg(**cu_rf_params)
cuml_mod = cuRFR_mg(**cu_rf_params, ignore_empty_partitions=True)
cuml_mod.fit(X_train_df, y_train_df)

cuml_mod_predict = cuml_mod.predict(X_test_df)
Expand All @@ -170,16 +173,17 @@ def test_rf_regression_dask_fil(partitions_per_worker,
@pytest.mark.parametrize('output_class', [True, False])
def test_rf_classification_dask_array(partitions_per_worker, client,
output_class):
n_workers = len(client.scheduler_info()['workers'])

X, y = make_classification(n_samples=10000, n_features=30,
X, y = make_classification(n_samples=n_workers * 2000, n_features=30,
n_clusters_per_class=1, n_informative=20,
random_state=123, n_classes=2)

X = X.astype(np.float32)
y = y.astype(np.int32)

X_train, X_test, y_train, y_test = \
train_test_split(X, y, test_size=1000)
train_test_split(X, y, test_size=n_workers * 400)

cu_rf_params = {
'n_estimators': 25,
Expand All @@ -204,15 +208,17 @@ def test_rf_classification_dask_array(partitions_per_worker, client,

@pytest.mark.parametrize('partitions_per_worker', [5])
def test_rf_regression_dask_cpu(partitions_per_worker, client):
X, y = make_regression(n_samples=10000, n_features=20,
n_workers = len(client.scheduler_info()['workers'])

X, y = make_regression(n_samples=n_workers * 2000, n_features=20,
n_informative=10, random_state=123)

X = X.astype(np.float32)
y = y.astype(np.float32)

X_train, X_test, y_train, y_test = train_test_split(X, y,
test_size=1000,
random_state=123)
X_train, X_test, y_train, y_test = \
train_test_split(X, y, test_size=n_workers * 400,
random_state=123)

cu_rf_params = {
'n_estimators': 50,
Expand Down Expand Up @@ -247,7 +253,9 @@ def test_rf_regression_dask_cpu(partitions_per_worker, client):
@pytest.mark.parametrize('partitions_per_worker', [5])
def test_rf_classification_dask_fil_predict_proba(partitions_per_worker,
client):
X, y = make_classification(n_samples=1000, n_features=30,
n_workers = len(client.scheduler_info()['workers'])

X, y = make_classification(n_samples=n_workers * 2700, n_features=30,
viclafargue marked this conversation as resolved.
Show resolved Hide resolved
n_clusters_per_class=1, n_informative=20,
random_state=123, n_classes=2)

Expand Down Expand Up @@ -288,8 +296,10 @@ def test_rf_classification_dask_fil_predict_proba(partitions_per_worker,

@pytest.mark.parametrize('model_type', ['classification', 'regression'])
def test_rf_concatenation_dask(client, model_type):
n_workers = len(client.scheduler_info()['workers'])

from cuml.fil.fil import TreeliteModel
X, y = make_classification(n_samples=1000, n_features=30,
X, y = make_classification(n_samples=n_workers * 200, n_features=30,
random_state=123, n_classes=2)

X = X.astype(np.float32)
Expand Down Expand Up @@ -358,6 +368,11 @@ def test_single_input(client, model_type, ignore_empty_partitions):
@pytest.mark.parametrize('n_estimators', [5, 10, 20])
@pytest.mark.parametrize('estimator_type', ['regression', 'classification'])
def test_rf_get_json(client, estimator_type, max_depth, n_estimators):
n_workers = len(client.scheduler_info()['workers'])
if n_estimators < n_workers:
err_msg = "n_estimators cannot be lower than number of dask workers"
pytest.xfail(err_msg)

X, y = make_classification(n_samples=350, n_features=20,
n_clusters_per_class=1, n_informative=10,
random_state=123, n_classes=2)
Expand Down Expand Up @@ -439,6 +454,8 @@ def predict_with_json_rf_regressor(rf, x):
@pytest.mark.parametrize('n_estimators', [5, 10, 20])
@pytest.mark.parametrize('detailed_text', [True, False])
def test_rf_get_text(client, n_estimators, detailed_text):
n_workers = len(client.scheduler_info()['workers'])

X, y = make_classification(n_samples=500, n_features=10,
n_clusters_per_class=1, n_informative=5,
random_state=94929, n_classes=2)
Expand All @@ -447,6 +464,15 @@ def test_rf_get_text(client, n_estimators, detailed_text):
y = y.astype(np.int32)
X, y = _prep_training_data(client, X, y, partitions_per_worker=2)

if n_estimators >= n_workers:
cu_rf_mg = cuRFC_mg(n_estimators=n_estimators,
ignore_empty_partitions=True)
else:
with pytest.raises(ValueError):
cu_rf_mg = cuRFC_mg(n_estimators=n_estimators,
ignore_empty_partitions=True)
return

cu_rf_mg = cuRFC_mg(n_estimators=n_estimators,
ignore_empty_partitions=True)
cu_rf_mg.fit(X, y)
Expand Down