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

Remove all use of DeviceQuantileDMatrix. (#8665)

Browse Source
This commit is contained in:
Jiaming Yuan
2023-01-17 00:04:10 +08:00
committed by GitHub
parent 0ae8df9a65
commit d6018eb4b9
10 changed files with 57 additions and 53 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
tests/python-gpu/test_device_quantile_dmatrix.py
View File

@@ -11,7 +11,7 @@ sys.path.append("tests/python")
import test_quantile_dmatrix as tqd
class TestDeviceQuantileDMatrix:
class TestQuantileDMatrix:
cputest = tqd.TestQuantileDMatrix()
@pytest.mark.skipif(**tm.no_cupy())
@@ -32,7 +32,7 @@ class TestDeviceQuantileDMatrix:
def test_dmatrix_cupy_init(self) -> None:
import cupy as cp
data = cp.random.randn(5, 5)
xgb.DeviceQuantileDMatrix(data, cp.ones(5, dtype=np.float64))
xgb.QuantileDMatrix(data, cp.ones(5, dtype=np.float64))
@pytest.mark.skipif(**tm.no_cupy())
@pytest.mark.parametrize(
@@ -85,7 +85,7 @@ class TestDeviceQuantileDMatrix:
fw = rng.randn(rows)
fw -= fw.min()
m = xgb.DeviceQuantileDMatrix(data=data, label=labels, feature_weights=fw)
m = xgb.QuantileDMatrix(data=data, label=labels, feature_weights=fw)
got_fw = m.get_float_info("feature_weights")
got_labels = m.get_label()

14
tests/python-gpu/test_from_cudf.py
View File

@@ -160,7 +160,7 @@ Arrow specification.'''
@pytest.mark.skipif(**tm.no_cudf())
def test_device_dmatrix_from_cudf(self):
_test_from_cudf(xgb.DeviceQuantileDMatrix)
_test_from_cudf(xgb.QuantileDMatrix)
@pytest.mark.skipif(**tm.no_cudf())
def test_cudf_training_simple_dmatrix(self):
@@ -168,7 +168,7 @@ Arrow specification.'''
@pytest.mark.skipif(**tm.no_cudf())
def test_cudf_training_device_dmatrix(self):
_test_cudf_training(xgb.DeviceQuantileDMatrix)
_test_cudf_training(xgb.QuantileDMatrix)
@pytest.mark.skipif(**tm.no_cudf())
def test_cudf_metainfo_simple_dmatrix(self):
@@ -176,7 +176,7 @@ Arrow specification.'''
@pytest.mark.skipif(**tm.no_cudf())
def test_cudf_metainfo_device_dmatrix(self):
_test_cudf_metainfo(xgb.DeviceQuantileDMatrix)
_test_cudf_metainfo(xgb.QuantileDMatrix)
@pytest.mark.skipif(**tm.no_cudf())
def test_cudf_categorical(self) -> None:
@@ -191,7 +191,7 @@ Arrow specification.'''
assert len(Xy.feature_types) == X.shape[1]
assert all(t == "c" for t in Xy.feature_types)
Xy = xgb.DeviceQuantileDMatrix(X, y, enable_categorical=True)
Xy = xgb.QuantileDMatrix(X, y, enable_categorical=True)
assert Xy.feature_types is not None
assert len(Xy.feature_types) == X.shape[1]
assert all(t == "c" for t in Xy.feature_types)
@@ -228,9 +228,9 @@ Arrow specification.'''
assert Xy.num_col() == 1
with pytest.raises(ValueError, match="enable_categorical"):
xgb.DeviceQuantileDMatrix(X, y)
xgb.QuantileDMatrix(X, y)
Xy = xgb.DeviceQuantileDMatrix(X, y, enable_categorical=True)
Xy = xgb.QuantileDMatrix(X, y, enable_categorical=True)
assert Xy.num_row() == 3
assert Xy.num_col() == 1
@@ -344,7 +344,7 @@ def test_from_cudf_iter(enable_categorical):
params = {"tree_method": "gpu_hist"}
# Use iterator
m_it = xgb.DeviceQuantileDMatrix(it, enable_categorical=enable_categorical)
m_it = xgb.QuantileDMatrix(it, enable_categorical=enable_categorical)
reg_with_it = xgb.train(params, m_it, num_boost_round=rounds)
X = it.as_array()

14
tests/python-gpu/test_from_cupy.py
View File

@@ -27,8 +27,8 @@ def dmatrix_from_cupy(input_type, DMatrixT, missing=np.NAN):
assert dtrain.num_col() == kCols
assert dtrain.num_row() == kRows
if DMatrixT is xgb.DeviceQuantileDMatrix:
# Slice is not supported by DeviceQuantileDMatrix
if DMatrixT is xgb.QuantileDMatrix:
# Slice is not supported by QuantileDMatrix
with pytest.raises(xgb.core.XGBoostError):
dtrain.slice(rindex=[0, 1, 2])
dtrain.slice(rindex=[0, 1, 2])
@@ -153,11 +153,11 @@ Arrow specification.'''
@pytest.mark.skipif(**tm.no_cupy())
def test_device_dmat_from_cupy(self):
_test_from_cupy(xgb.DeviceQuantileDMatrix)
_test_from_cupy(xgb.QuantileDMatrix)
@pytest.mark.skipif(**tm.no_cupy())
def test_cupy_training_device_dmat(self):
_test_cupy_training(xgb.DeviceQuantileDMatrix)
_test_cupy_training(xgb.QuantileDMatrix)
@pytest.mark.skipif(**tm.no_cupy())
def test_cupy_training_simple_dmat(self):
@@ -169,7 +169,7 @@ Arrow specification.'''
@pytest.mark.skipif(**tm.no_cupy())
def test_cupy_metainfo_device_dmat(self):
_test_cupy_metainfo(xgb.DeviceQuantileDMatrix)
_test_cupy_metainfo(xgb.QuantileDMatrix)
@pytest.mark.skipif(**tm.no_cupy())
def test_dlpack_simple_dmat(self):
@@ -196,7 +196,7 @@ Arrow specification.'''
import cupy as cp
n = 100
X = cp.random.random((n, 2))
m = xgb.DeviceQuantileDMatrix(X.toDlpack())
m = xgb.QuantileDMatrix(X.toDlpack())
with pytest.raises(xgb.core.XGBoostError):
m.slice(rindex=[0, 1, 2])
@@ -222,7 +222,7 @@ Arrow specification.'''
import cupy as cp
cp.cuda.runtime.setDevice(0)
dtrain = dmatrix_from_cupy(
np.float32, xgb.DeviceQuantileDMatrix, np.nan)
np.float32, xgb.QuantileDMatrix, np.nan)
with pytest.raises(xgb.core.XGBoostError):
xgb.train(
{'tree_method': 'gpu_hist', 'gpu_id': 1}, dtrain, num_boost_round=10

2
tests/python-gpu/test_large_input.py
View File

@@ -17,7 +17,7 @@ def test_large_input():
assert (np.log2(m * n) > 31)
X = cp.ones((m, n), dtype=np.float32)
y = cp.ones(m)
dmat = xgb.DeviceQuantileDMatrix(X, y)
dmat = xgb.QuantileDMatrix(X, y)
booster = xgb.train({"tree_method": "gpu_hist", "max_depth": 1}, dmat, 1)
del y
booster.inplace_predict(X)

2
tests/python/test_updaters.py
View File

@@ -173,7 +173,7 @@ class TestTreeMethod:
X, y = cp.array(X), cp.array(y)
with pytest.raises(ValueError):
Xy = xgb.DeviceQuantileDMatrix(X, y, feature_types=["c"] * 10)
Xy = xgb.QuantileDMatrix(X, y, feature_types=["c"] * 10)
def test_invalid_category(self) -> None:
self.run_invalid_category("approx")

16
tests/test_distributed/test_gpu_with_dask/test_gpu_with_dask.py
View File

@@ -135,7 +135,7 @@ def run_with_dask_array(DMatrixT: Type, client: Client) -> None:
def to_cp(x: Any, DMatrixT: Type) -> Any:
import cupy
if isinstance(x, np.ndarray) and DMatrixT is dxgb.DaskDeviceQuantileDMatrix:
if isinstance(x, np.ndarray) and DMatrixT is dxgb.DaskQuantileDMatrix:
X = cupy.array(x)
else:
X = x
@@ -169,7 +169,7 @@ def run_gpu_hist(
else:
w = None
if DMatrixT is dxgb.DaskDeviceQuantileDMatrix:
if DMatrixT is dxgb.DaskQuantileDMatrix:
m = DMatrixT(
client, data=X, label=y, weight=w, max_bin=params.get("max_bin", 256)
)
@@ -227,7 +227,7 @@ class TestDistributedGPU:
@pytest.mark.skipif(**tm.no_dask_cudf())
def test_dask_dataframe(self, local_cuda_client: Client) -> None:
run_with_dask_dataframe(dxgb.DaskDMatrix, local_cuda_client)
run_with_dask_dataframe(dxgb.DaskDeviceQuantileDMatrix, local_cuda_client)
run_with_dask_dataframe(dxgb.DaskQuantileDMatrix, local_cuda_client)
@pytest.mark.skipif(**tm.no_dask_cudf())
def test_categorical(self, local_cuda_client: Client) -> None:
@@ -245,7 +245,7 @@ class TestDistributedGPU:
num_rounds=strategies.integers(1, 20),
dataset=tm.dataset_strategy,
dmatrix_type=strategies.sampled_from(
[dxgb.DaskDMatrix, dxgb.DaskDeviceQuantileDMatrix]
[dxgb.DaskDMatrix, dxgb.DaskQuantileDMatrix]
),
)
@settings(
@@ -268,7 +268,7 @@ class TestDistributedGPU:
@pytest.mark.skipif(**tm.no_cupy())
def test_dask_array(self, local_cuda_client: Client) -> None:
run_with_dask_array(dxgb.DaskDMatrix, local_cuda_client)
run_with_dask_array(dxgb.DaskDeviceQuantileDMatrix, local_cuda_client)
run_with_dask_array(dxgb.DaskQuantileDMatrix, local_cuda_client)
@pytest.mark.skipif(**tm.no_cupy())
def test_early_stopping(self, local_cuda_client: Client) -> None:
@@ -357,7 +357,7 @@ class TestDistributedGPU:
)
X = ddf[ddf.columns.difference(["y"])]
y = ddf[["y"]]
dtrain = dxgb.DaskDeviceQuantileDMatrix(local_cuda_client, X, y)
dtrain = dxgb.DaskQuantileDMatrix(local_cuda_client, X, y)
bst_empty = xgb.dask.train(
local_cuda_client, parameters, dtrain, evals=[(dtrain, "train")]
)
@@ -369,7 +369,7 @@ class TestDistributedGPU:
)
X = ddf[ddf.columns.difference(["y"])]
y = ddf[["y"]]
dtrain = dxgb.DaskDeviceQuantileDMatrix(local_cuda_client, X, y)
dtrain = dxgb.DaskQuantileDMatrix(local_cuda_client, X, y)
bst = xgb.dask.train(
local_cuda_client, parameters, dtrain, evals=[(dtrain, "train")]
)
@@ -546,7 +546,7 @@ async def run_from_dask_array_asyncio(scheduler_address: str) -> dxgb.TrainRetur
X = X.map_blocks(cp.array)
y = y.map_blocks(cp.array)
m = await xgb.dask.DaskDeviceQuantileDMatrix(client, X, y)
m = await xgb.dask.DaskQuantileDMatrix(client, X, y)
output = await xgb.dask.train(client, {"tree_method": "gpu_hist"}, dtrain=m)
with_m = await xgb.dask.predict(client, output, m)