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

Enforce correct data shape. (#5191)

Browse Source 
* Fix syncing DMatrix columns.
* notes for tree method.
* Enable feature validation for all interfaces except for jvm.
* Better tests for boosting from predictions.
* Disable validation on JVM.
This commit is contained in:
Jiaming Yuan
2020-01-13 15:48:17 +08:00
committed by GitHub
parent 8cbcc53ccb
commit 7b65698187
14 changed files with 108 additions and 60 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
tests/cpp/test_learner.cc
View File

@@ -91,7 +91,6 @@ TEST(Learner, CheckGroup) {
}
TEST(Learner, SLOW_CheckMultiBatch) {
using Arg = std::pair<std::string, std::string>;
// Create sufficiently large data to make two row pages
dmlc::TemporaryDirectory tempdir;
const std::string tmp_file = tempdir.path + "/big.libsvm";
@@ -107,7 +106,7 @@ TEST(Learner, SLOW_CheckMultiBatch) {
dmat->Info().SetInfo("label", labels.data(), DataType::kFloat32, num_row);
std::vector<std::shared_ptr<DMatrix>> mat{dmat};
auto learner = std::unique_ptr<Learner>(Learner::Create(mat));
learner->SetParams({Arg{"objective", "binary:logistic"}, Arg{"verbosity", "3"}});
learner->SetParams(Args{{"objective", "binary:logistic"}});
learner->UpdateOneIter(0, dmat.get());
}

1
tests/python-gpu/test_gpu_pickling.py
View File

@@ -6,7 +6,6 @@ import subprocess
import os
import json
import pytest
import copy
import xgboost as xgb
from xgboost import XGBClassifier

12
tests/python-gpu/test_gpu_prediction.py
View File

@@ -13,6 +13,10 @@ class TestGPUPredict(unittest.TestCase):
np.random.seed(1)
test_num_rows = [10, 1000, 5000]
test_num_cols = [10, 50, 500]
# This test passes for tree_method=gpu_hist and tree_method=exact. but
# for `hist` and `approx` the floating point error accumulates faster
# and fails even tol is set to 1e-4. For `hist`, the mismatching rate
# with 5000 rows is 0.04.
for num_rows in test_num_rows:
for num_cols in test_num_cols:
dtrain = xgb.DMatrix(np.random.randn(num_rows, num_cols),
@@ -27,7 +31,7 @@ class TestGPUPredict(unittest.TestCase):
"objective": "binary:logistic",
"predictor": "gpu_predictor",
'eval_metric': 'auc',
'verbosity': '3'
'tree_method': 'gpu_hist'
}
bst = xgb.train(param, dtrain, iterations, evals=watchlist,
evals_result=res)
@@ -43,11 +47,11 @@ class TestGPUPredict(unittest.TestCase):
cpu_pred_val = bst_cpu.predict(dval, output_margin=True)
np.testing.assert_allclose(cpu_pred_train, gpu_pred_train,
rtol=1e-3)
rtol=1e-6)
np.testing.assert_allclose(cpu_pred_val, gpu_pred_val,
rtol=1e-3)
rtol=1e-6)
np.testing.assert_allclose(cpu_pred_test, gpu_pred_test,
rtol=1e-3)
rtol=1e-6)
def non_decreasing(self, L):
return all((x - y) < 0.001 for x, y in zip(L, L[1:]))

11
tests/python-gpu/test_gpu_with_sklearn.py
View File

@@ -2,9 +2,11 @@ import xgboost as xgb
import pytest
import sys
import numpy as np
import unittest
sys.path.append("tests/python")
import testing as tm
import testing as tm # noqa
import test_with_sklearn as twskl # noqa
pytestmark = pytest.mark.skipif(**tm.no_sklearn())
@@ -29,3 +31,10 @@ def test_gpu_binary_classification():
err = sum(1 for i in range(len(preds))
if int(preds[i] > 0.5) != labels[i]) / float(len(preds))
assert err < 0.1
class TestGPUBoostFromPrediction(unittest.TestCase):
cpu_test = twskl.TestBoostFromPrediction()
def test_boost_from_prediction_gpu_hist(self):
self.cpu_test.run_boost_from_prediction('gpu_hist')

49
tests/python/test_with_sklearn.py
View File

@@ -5,6 +5,7 @@ import tempfile
import os
import shutil
import pytest
import unittest
rng = np.random.RandomState(1994)
@@ -697,21 +698,37 @@ def test_XGBClassifier_resume():
assert log_loss1 > log_loss2
def test_boost_from_prediction():
from sklearn.datasets import load_breast_cancer
X, y = load_breast_cancer(return_X_y=True)
model_0 = xgb.XGBClassifier(
learning_rate=0.3, random_state=0, n_estimators=4)
model_0.fit(X=X, y=y)
margin = model_0.predict(X, output_margin=True)
class TestBoostFromPrediction(unittest.TestCase):
def run_boost_from_prediction(self, tree_method):
from sklearn.datasets import load_breast_cancer
X, y = load_breast_cancer(return_X_y=True)
model_0 = xgb.XGBClassifier(
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.XGBClassifier(
learning_rate=0.3, random_state=0, n_estimators=4)
model_1.fit(X=X, y=y, base_margin=margin)
predictions_1 = model_1.predict(X, base_margin=margin)
model_1 = xgb.XGBClassifier(
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.XGBClassifier(
learning_rate=0.3, random_state=0, n_estimators=8)
cls_2.fit(X=X, y=y)
predictions_2 = cls_2.predict(X, base_margin=margin)
assert np.all(predictions_1 == predictions_2)
cls_2 = xgb.XGBClassifier(
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 == predictions_2)
@pytest.mark.skipif(**tm.no_sklearn())
def test_boost_from_prediction_hist(self):
self.run_boost_from_prediction('hist')
@pytest.mark.skipif(**tm.no_sklearn())
def test_boost_from_prediction_approx(self):
self.run_boost_from_prediction('approx')
@pytest.mark.skipif(**tm.no_sklearn())
def test_boost_from_prediction_exact(self):
self.run_boost_from_prediction('exact')