Hendrik/xgboost
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Implement max_cat_threshold for CPU. (#7957)

Browse Source
This commit is contained in:
Jiaming Yuan
2022-06-04 11:02:46 +08:00
committed by GitHub
parent 78694405a6
commit b90c6d25e8
8 changed files with 177 additions and 20 deletions
Download Patch File Download Diff File Expand all files Collapse all files

62
tests/python/test_updaters.py
View File

@@ -31,6 +31,14 @@ hist_parameter_strategy = strategies.fixed_dictionaries({
x['max_depth'] > 0 or x['grow_policy'] == 'lossguide'))
cat_parameter_strategy = strategies.fixed_dictionaries(
{
"max_cat_to_onehot": strategies.integers(1, 128),
"max_cat_threshold": strategies.integers(1, 128),
}
)
def train_result(param, dmat, num_rounds):
result = {}
xgb.train(param, dmat, num_rounds, [(dmat, 'train')], verbose_eval=False,
@@ -253,7 +261,7 @@ class TestTreeMethod:
# Test with partition-based split
run(self.USE_PART)
def run_categorical_basic(self, rows, cols, rounds, cats, tree_method):
def run_categorical_ohe(self, rows, cols, rounds, cats, tree_method):
onehot, label = tm.make_categorical(rows, cols, cats, True)
cat, _ = tm.make_categorical(rows, cols, cats, False)
@@ -328,9 +336,55 @@ class TestTreeMethod:
strategies.integers(1, 2), strategies.integers(4, 7))
@settings(deadline=None, print_blob=True)
@pytest.mark.skipif(**tm.no_pandas())
def test_categorical(self, rows, cols, rounds, cats):
self.run_categorical_basic(rows, cols, rounds, cats, "approx")
self.run_categorical_basic(rows, cols, rounds, cats, "hist")
def test_categorical_ohe(self, rows, cols, rounds, cats):
self.run_categorical_ohe(rows, cols, rounds, cats, "approx")
self.run_categorical_ohe(rows, cols, rounds, cats, "hist")
@given(
tm.categorical_dataset_strategy,
exact_parameter_strategy,
hist_parameter_strategy,
cat_parameter_strategy,
strategies.integers(4, 32),
strategies.sampled_from(["hist", "approx"]),
)
@settings(deadline=None, print_blob=True)
@pytest.mark.skipif(**tm.no_pandas())
def test_categorical(
self,
dataset: tm.TestDataset,
exact_parameters: Dict[str, Any],
hist_parameters: Dict[str, Any],
cat_parameters: Dict[str, Any],
n_rounds: int,
tree_method: str,
) -> None:
cat_parameters.update(exact_parameters)
cat_parameters.update(hist_parameters)
cat_parameters["tree_method"] = tree_method
results = train_result(cat_parameters, dataset.get_dmat(), n_rounds)
tm.non_increasing(results["train"]["rmse"])
@given(
hist_parameter_strategy,
cat_parameter_strategy,
strategies.sampled_from(["hist", "approx"]),
)
@settings(deadline=None, print_blob=True)
def test_categorical_ames_housing(
self,
hist_parameters: Dict[str, Any],
cat_parameters: Dict[str, Any],
tree_method: str,
) -> None:
cat_parameters.update(hist_parameters)
dataset = tm.TestDataset(
"ames_housing", tm.get_ames_housing, "reg:squarederror", "rmse"
)
cat_parameters["tree_method"] = tree_method
results = train_result(cat_parameters, dataset.get_dmat(), 16)
tm.non_increasing(results["train"]["rmse"])
@given(
strategies.integers(10, 400),

82
tests/python/testing.py
View File

@@ -214,7 +214,9 @@ class TestDataset:
return params_in
def get_dmat(self):
return xgb.DMatrix(self.X, self.y, self.w, base_margin=self.margin)
return xgb.DMatrix(
self.X, self.y, self.w, base_margin=self.margin, enable_categorical=True
)
def get_device_dmat(self):
w = None if self.w is None else cp.array(self.w)
@@ -277,6 +279,48 @@ def get_sparse():
return X, y
@memory.cache
def get_ames_housing():
"""
Number of samples: 1460
Number of features: 20
Number of categorical features: 10
Number of numerical features: 10
"""
from sklearn.datasets import fetch_openml
X, y = fetch_openml(data_id=42165, as_frame=True, return_X_y=True)
categorical_columns_subset: list[str] = [
"BldgType", # 5 cats, no nan
"GarageFinish", # 3 cats, nan
"LotConfig", # 5 cats, no nan
"Functional", # 7 cats, no nan
"MasVnrType", # 4 cats, nan
"HouseStyle", # 8 cats, no nan
"FireplaceQu", # 5 cats, nan
"ExterCond", # 5 cats, no nan
"ExterQual", # 4 cats, no nan
"PoolQC", # 3 cats, nan
]
numerical_columns_subset: list[str] = [
"3SsnPorch",
"Fireplaces",
"BsmtHalfBath",
"HalfBath",
"GarageCars",
"TotRmsAbvGrd",
"BsmtFinSF1",
"BsmtFinSF2",
"GrLivArea",
"ScreenPorch",
]
X = X[categorical_columns_subset + numerical_columns_subset]
X[categorical_columns_subset] = X[categorical_columns_subset].astype("category")
return X, y
@memory.cache
def get_mq2008(dpath):
from sklearn.datasets import load_svmlight_files
@@ -329,7 +373,6 @@ def make_categorical(
for i in range(n_features):
index = rng.randint(low=0, high=n_samples-1, size=int(n_samples * sparsity))
df.iloc[index, i] = np.NaN
assert df.iloc[:, i].isnull().values.any()
assert n_categories == np.unique(df.dtypes[i].categories).size
if onehot:
@@ -337,6 +380,41 @@ def make_categorical(
return df, label
def _cat_sampled_from():
@strategies.composite
def _make_cat(draw):
n_samples = draw(strategies.integers(2, 512))
n_features = draw(strategies.integers(1, 4))
n_cats = draw(strategies.integers(1, 128))
sparsity = draw(
strategies.floats(
min_value=0,
max_value=1,
allow_nan=False,
allow_infinity=False,
allow_subnormal=False,
)
)
return n_samples, n_features, n_cats, sparsity
def _build(args):
n_samples = args[0]
n_features = args[1]
n_cats = args[2]
sparsity = args[3]
return TestDataset(
f"{n_samples}x{n_features}-{n_cats}-{sparsity}",
lambda: make_categorical(n_samples, n_features, n_cats, False, sparsity),
"reg:squarederror",
"rmse",
)
return _make_cat().map(_build)
categorical_dataset_strategy = _cat_sampled_from()
@memory.cache
def make_sparse_regression(
n_samples: int, n_features: int, sparsity: float, as_dense: bool