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

Replaced std::vector-based interfaces with HostDeviceVector-based interfaces. (#3116)

Browse Source 
* Replaced std::vector-based interfaces with HostDeviceVector-based interfaces.

- replacement was performed in the learner, boosters, predictors,
  updaters, and objective functions
- only interfaces used in training were replaced;
  interfaces like PredictInstance() still use std::vector
- refactoring necessary for replacement of interfaces was also performed,
  such as using HostDeviceVector in prediction cache

* HostDeviceVector-based interfaces for custom objective function example plugin.
This commit is contained in:
Andrew V. Adinetz
2018-02-28 01:00:04 +01:00
committed by Rory Mitchell
parent 11bfa8584d
commit d5992dd881
38 changed files with 371 additions and 519 deletions
Download Patch File Download Diff File Expand all files Collapse all files

9
tests/cpp/helpers.cc
View File

@@ -38,10 +38,13 @@ void CheckObjFunction(xgboost::ObjFunction * obj,
info.labels = labels;
info.weights = weights;
std::vector<xgboost::bst_gpair> gpair;
obj->GetGradient(preds, info, 1, &gpair);
xgboost::HostDeviceVector<xgboost::bst_float> in_preds(preds);
ASSERT_EQ(gpair.size(), preds.size());
xgboost::HostDeviceVector<xgboost::bst_gpair> out_gpair;
obj->GetGradient(&in_preds, info, 1, &out_gpair);
std::vector<xgboost::bst_gpair>& gpair = out_gpair.data_h();
ASSERT_EQ(gpair.size(), in_preds.size());
for (int i = 0; i < static_cast<int>(gpair.size()); ++i) {
EXPECT_NEAR(gpair[i].GetGrad(), out_grad[i], 0.01)
<< "Unexpected grad for pred=" << preds[i] << " label=" << labels[i]

28
tests/cpp/objective/test_regression_obj.cc
View File

@@ -46,10 +46,11 @@ TEST(Objective, LogisticRegressionBasic) {
<< "Expected error when base_score not in range [0,1f] for LogisticRegression";
// test PredTransform
std::vector<xgboost::bst_float> preds = {0, 0.1f, 0.5f, 0.9f, 1};
xgboost::HostDeviceVector<xgboost::bst_float> io_preds = {0, 0.1f, 0.5f, 0.9f, 1};
std::vector<xgboost::bst_float> out_preds = {0.5f, 0.524f, 0.622f, 0.710f, 0.731f};
obj->PredTransform(&preds);
for (int i = 0; i < static_cast<int>(preds.size()); ++i) {
obj->PredTransform(&io_preds);
auto& preds = io_preds.data_h();
for (int i = 0; i < static_cast<int>(io_preds.size()); ++i) {
EXPECT_NEAR(preds[i], out_preds[i], 0.01f);
}
}
@@ -94,10 +95,11 @@ TEST(Objective, PoissonRegressionBasic) {
EXPECT_NEAR(obj->ProbToMargin(0.9f), -0.10f, 0.01f);
// test PredTransform
std::vector<xgboost::bst_float> preds = {0, 0.1f, 0.5f, 0.9f, 1};
xgboost::HostDeviceVector<xgboost::bst_float> io_preds = {0, 0.1f, 0.5f, 0.9f, 1};
std::vector<xgboost::bst_float> out_preds = {1, 1.10f, 1.64f, 2.45f, 2.71f};
obj->PredTransform(&preds);
for (int i = 0; i < static_cast<int>(preds.size()); ++i) {
obj->PredTransform(&io_preds);
auto& preds = io_preds.data_h();
for (int i = 0; i < static_cast<int>(io_preds.size()); ++i) {
EXPECT_NEAR(preds[i], out_preds[i], 0.01f);
}
}
@@ -129,10 +131,11 @@ TEST(Objective, GammaRegressionBasic) {
EXPECT_NEAR(obj->ProbToMargin(0.9f), -0.10f, 0.01f);
// test PredTransform
std::vector<xgboost::bst_float> preds = {0, 0.1f, 0.5f, 0.9f, 1};
xgboost::HostDeviceVector<xgboost::bst_float> io_preds = {0, 0.1f, 0.5f, 0.9f, 1};
std::vector<xgboost::bst_float> out_preds = {1, 1.10f, 1.64f, 2.45f, 2.71f};
obj->PredTransform(&preds);
for (int i = 0; i < static_cast<int>(preds.size()); ++i) {
obj->PredTransform(&io_preds);
auto& preds = io_preds.data_h();
for (int i = 0; i < static_cast<int>(io_preds.size()); ++i) {
EXPECT_NEAR(preds[i], out_preds[i], 0.01f);
}
}
@@ -165,10 +168,11 @@ TEST(Objective, TweedieRegressionBasic) {
EXPECT_NEAR(obj->ProbToMargin(0.9f), 0.89f, 0.01f);
// test PredTransform
std::vector<xgboost::bst_float> preds = {0, 0.1f, 0.5f, 0.9f, 1};
xgboost::HostDeviceVector<xgboost::bst_float> io_preds = {0, 0.1f, 0.5f, 0.9f, 1};
std::vector<xgboost::bst_float> out_preds = {1, 1.10f, 1.64f, 2.45f, 2.71f};
obj->PredTransform(&preds);
for (int i = 0; i < static_cast<int>(preds.size()); ++i) {
obj->PredTransform(&io_preds);
auto& preds = io_preds.data_h();
for (int i = 0; i < static_cast<int>(io_preds.size()); ++i) {
EXPECT_NEAR(preds[i], out_preds[i], 0.01f);
}
}

7
tests/cpp/objective/test_regression_obj_gpu.cu
View File

@@ -48,10 +48,11 @@ TEST(Objective, GPULogisticRegressionBasic) {
<< "Expected error when base_score not in range [0,1f] for LogisticRegression";
// test PredTransform
std::vector<xgboost::bst_float> preds = {0, 0.1f, 0.5f, 0.9f, 1};
xgboost::HostDeviceVector<xgboost::bst_float> io_preds = {0, 0.1f, 0.5f, 0.9f, 1};
std::vector<xgboost::bst_float> out_preds = {0.5f, 0.524f, 0.622f, 0.710f, 0.731f};
obj->PredTransform(&preds);
for (int i = 0; i < static_cast<int>(preds.size()); ++i) {
obj->PredTransform(&io_preds);
auto& preds = io_preds.data_h();
for (int i = 0; i < static_cast<int>(io_preds.size()); ++i) {
EXPECT_NEAR(preds[i], out_preds[i], 0.01f);
}
}

5
tests/cpp/predictor/test_cpu_predictor.cc
View File

@@ -24,10 +24,11 @@ TEST(cpu_predictor, Test) {
auto dmat = CreateDMatrix(n_row, n_col, 0);
// Test predict batch
std::vector<float> out_predictions;
HostDeviceVector<float> out_predictions;
cpu_predictor->PredictBatch(dmat.get(), &out_predictions, model, 0);
std::vector<float>& out_predictions_h = out_predictions.data_h();
for (int i = 0; i < out_predictions.size(); i++) {
ASSERT_EQ(out_predictions[i], 1.5);
ASSERT_EQ(out_predictions_h[i], 1.5);
}
// Test predict instance

8
tests/cpp/predictor/test_gpu_predictor.cu
View File

@@ -33,13 +33,15 @@ TEST(gpu_predictor, Test) {
auto dmat = CreateDMatrix(n_row, n_col, 0);
// Test predict batch
std::vector<float> gpu_out_predictions;
std::vector<float> cpu_out_predictions;
HostDeviceVector<float> gpu_out_predictions;
HostDeviceVector<float> cpu_out_predictions;
gpu_predictor->PredictBatch(dmat.get(), &gpu_out_predictions, model, 0);
cpu_predictor->PredictBatch(dmat.get(), &cpu_out_predictions, model, 0);
std::vector<float>& gpu_out_predictions_h = gpu_out_predictions.data_h();
std::vector<float>& cpu_out_predictions_h = cpu_out_predictions.data_h();
float abs_tolerance = 0.001;
for (int i = 0; i < gpu_out_predictions.size(); i++) {
ASSERT_LT(std::abs(gpu_out_predictions[i] - cpu_out_predictions[i]),
ASSERT_LT(std::abs(gpu_out_predictions_h[i] - cpu_out_predictions_h[i]),
abs_tolerance);
}
// Test predict instance