Skip to content

Commit

Permalink
Update failing MNMG tests (#3348)
Browse files Browse the repository at this point in the history 
Partially answers #3354.
The PR updates some of the failing MNMG tests so that nightly testing finally pass. This PR addresses MNMG RF and MNMG KNN tests failures.

Authors:
  - Victor Lafargue (@viclafargue)

Approvers:
  - John Zedlewski (@JohnZed)

URL: #3348
  • Loading branch information
@viclafargue
viclafargue authored Feb 8, 2021
1 parent c9c8619 commit c1a7447
Show file tree
Hide file tree
Showing 2 changed files with 45 additions and 19 deletions.
2 changes: 1 addition & 1 deletion python/cuml/test/dask/test_nearest_neighbors.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -249,7 +249,7 @@ 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
62 changes: 44 additions & 18 deletions python/cuml/test/dask/test_random_forest.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -78,18 +78,19 @@ 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 @@ -101,7 +102,7 @@ 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 @@ -114,7 +115,7 @@ 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 @@ -124,16 +125,18 @@ 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 @@ -158,7 +161,7 @@ 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 @@ -173,16 +176,17 @@ 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 @@ -207,15 +211,17 @@ 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 @@ -250,15 +256,17 @@ 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 * 1500, 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=100, random_state=123)
train_test_split(X, y, test_size=n_workers * 150, random_state=123)

cu_rf_params = {'n_bins': 16, 'n_streams': 1,
'n_estimators': 40, 'max_depth': 16
Expand Down Expand Up @@ -291,8 +299,10 @@ 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 @@ -361,6 +371,11 @@ 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 @@ -483,6 +498,8 @@ def test_rf_get_combined_model_right_aftter_fit(client, estimator_type):
@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 @@ -491,6 +508,15 @@ 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

0 comments on commit c1a7447

Please sign in to comment.