Skip to content

GH-16590 Accept zero as offset for xgb models trained with offset #16605

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
Mathanraj-Sharma merged 1 commit into from
May 7, 2025
Merged

GH-16590 Accept zero as offset for xgb models trained with offset #16605

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion h2o-algos/src/main/java/hex/generic/GenericModel.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -313,7 +313,7 @@ protected double[] score0(double[] data, double[] preds) {

@Override
protected double[] score0(double[] data, double[] preds, double offset) {
if (offset == 0) // MOJO doesn't like when score0 is called with 0 offset for problems that were trained without offset
if (!_output.hasOffset() && offset == 0) // MOJO doesn't like when score0 is called with 0 offset for problems that were trained without offset
return score0(data, preds);
else
return genModel().score0(data, offset, preds);
Expand Down
244 changes: 174 additions & 70 deletions ...extensions/xgboost/src/test/java/hex/genmodel/algos/xgboost/XGBoostJavaMojoModelTest.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -7,93 +7,197 @@
import hex.genmodel.algos.tree.SharedTreeGraph;
import hex.genmodel.algos.tree.SharedTreeMojoModel;
import hex.genmodel.algos.tree.SharedTreeNode;
import hex.genmodel.algos.tree.SharedTreeSubgraph;
import hex.genmodel.easy.EasyPredictModelWrapper;
import hex.genmodel.easy.RowData;
import hex.genmodel.easy.prediction.RegressionModelPrediction;
import org.junit.Test;

import java.io.*;
import java.nio.file.Paths;

import static org.junit.Assert.*;

public class XGBoostJavaMojoModelTest {

@Test
public void testObjFunction() { // make sure we have implementation for all supported obj functions
for (XGBoostMojoModel.ObjectiveType type : XGBoostMojoModel.ObjectiveType.values()) {
assertNotNull(type.getId());
assertFalse(type.getId().isEmpty());
// check we have an implementation of ObjFunction
assertNotNull(XGBoostJavaMojoModel.getObjFunction(type.getId()));
@Test
public void testObjFunction() { // make sure we have implementation for all supported obj functions
for (XGBoostMojoModel.ObjectiveType type : XGBoostMojoModel.ObjectiveType.values()) {
assertNotNull(type.getId());
assertFalse(type.getId().isEmpty());
// check we have an implementation of ObjFunction
assertNotNull(XGBoostJavaMojoModel.getObjFunction(type.getId()));
}
}
}

@Test
public void testPredictContributionsSerialization() throws Exception {
MojoReaderBackend readerBackend = MojoReaderBackendFactory.createReaderBackend(
XGBoostJavaMojoModelTest.class.getResource("xgboost_java.zip"),
MojoReaderBackendFactory.CachingStrategy.MEMORY);
XGBoostJavaMojoModel mojo = (XGBoostJavaMojoModel) MojoModel.load(readerBackend);
PredictContributions pc = mojo.makeContributionsPredictor();
assertNotNull(pc);
assertTrue(deserialize(serialize(pc)) instanceof PredictContributions);
}

@Test
public void testLeafNodeAssignments() throws Exception {
MojoReaderBackend readerBackend = MojoReaderBackendFactory.createReaderBackend(
getClass().getResource("xgboost_java.zip"),
MojoReaderBackendFactory.CachingStrategy.MEMORY);
XGBoostJavaMojoModel mojo = (XGBoostJavaMojoModel) MojoModel.load(readerBackend);
double[] doubles = new double[]{1, 2, 3, 4, 5, 6, 7};
SharedTreeMojoModel.LeafNodeAssignments res = mojo.getLeafNodeAssignments(doubles);
assertNotNull(res._nodeIds);
assertNotNull(res._paths);
String[] paths = mojo.getDecisionPath(doubles);
assertArrayEquals(paths, res._paths);
RowData data = new RowData();
for (int i = 0; i< doubles.length; i++) data.put(mojo._names[i], doubles[i]);
EasyPredictModelWrapper wrapper = new EasyPredictModelWrapper(
new EasyPredictModelWrapper.Config().setModel(mojo).setEnableLeafAssignment(true)
);
RegressionModelPrediction res2 = (RegressionModelPrediction) wrapper.predict(data);
assertNotNull(res2.leafNodeAssignmentIds);
assertNotNull(res2.leafNodeAssignments);
assertArrayEquals(res._nodeIds, res2.leafNodeAssignmentIds);
assertArrayEquals(res._paths, res2.leafNodeAssignments);
}

@Test
public void testConvertWithWeights() throws Exception {
MojoReaderBackend readerBackend = MojoReaderBackendFactory.createReaderBackend(
XGBoostJavaMojoModelTest.class.getResource("xgboost_java.zip"),
MojoReaderBackendFactory.CachingStrategy.MEMORY);
XGBoostJavaMojoModel mojo = (XGBoostJavaMojoModel) MojoModel.load(readerBackend);
SharedTreeGraph graph = mojo.convert(0, null);
int expectedWeight = 380; // prostate dataset, 380 rows
assertEquals(graph.subgraphArray.get(0).rootNode.getWeight(), expectedWeight, 0);
double actualWeight = 0;
for (SharedTreeNode node : graph.subgraphArray.get(0).nodesArray) {
actualWeight += node.getWeight();

@Test
public void testPredictContributionsSerialization() throws Exception {
MojoReaderBackend readerBackend = MojoReaderBackendFactory.createReaderBackend(
XGBoostJavaMojoModelTest.class.getResource("xgboost_java.zip"),
MojoReaderBackendFactory.CachingStrategy.MEMORY);
XGBoostJavaMojoModel mojo = (XGBoostJavaMojoModel) MojoModel.load(readerBackend);
PredictContributions pc = mojo.makeContributionsPredictor();
assertNotNull(pc);
assertTrue(deserialize(serialize(pc)) instanceof PredictContributions);
}

@Test
public void testLeafNodeAssignments() throws Exception {
MojoReaderBackend readerBackend = MojoReaderBackendFactory.createReaderBackend(
getClass().getResource("xgboost_java.zip"),
MojoReaderBackendFactory.CachingStrategy.MEMORY);
XGBoostJavaMojoModel mojo = (XGBoostJavaMojoModel) MojoModel.load(readerBackend);
double[] doubles = new double[]{1, 2, 3, 4, 5, 6, 7};
SharedTreeMojoModel.LeafNodeAssignments res = mojo.getLeafNodeAssignments(doubles);
assertNotNull(res._nodeIds);
assertNotNull(res._paths);
String[] paths = mojo.getDecisionPath(doubles);
assertArrayEquals(paths, res._paths);
RowData data = new RowData();
for (int i = 0; i < doubles.length; i++) data.put(mojo._names[i], doubles[i]);
EasyPredictModelWrapper wrapper = new EasyPredictModelWrapper(
new EasyPredictModelWrapper.Config().setModel(mojo).setEnableLeafAssignment(true)
);
RegressionModelPrediction res2 = (RegressionModelPrediction) wrapper.predict(data);
assertNotNull(res2.leafNodeAssignmentIds);
assertNotNull(res2.leafNodeAssignments);
assertArrayEquals(res._nodeIds, res2.leafNodeAssignmentIds);
assertArrayEquals(res._paths, res2.leafNodeAssignments);
}

@Test
public void testConvertWithWeights() throws Exception {
MojoReaderBackend readerBackend = MojoReaderBackendFactory.createReaderBackend(
XGBoostJavaMojoModelTest.class.getResource("xgboost_java.zip"),
MojoReaderBackendFactory.CachingStrategy.MEMORY);
XGBoostJavaMojoModel mojo = (XGBoostJavaMojoModel) MojoModel.load(readerBackend);
SharedTreeGraph graph = mojo.convert(0, null);
int expectedWeight = 380; // prostate dataset, 380 rows
assertEquals(graph.subgraphArray.get(0).rootNode.getWeight(), expectedWeight, 0);
double actualWeight = 0;
for (SharedTreeNode node : graph.subgraphArray.get(0).nodesArray) {
actualWeight += node.getWeight();
}
assertEquals(expectedWeight, actualWeight, 0);
}
assertEquals(expectedWeight, actualWeight, 0);
}

/*
Test for XGBoost model trained with offset and predicting with none zero offset.
*/
@Test
public void testXGBWithOffset_NoneZeroOffset() throws Exception {
String mojofile = String.valueOf(
Paths.get(
XGBoostJavaMojoModelTest.class.getClassLoader().getResource("hex/genmodel/algos/xgboost/XGBoostWithOffset.zip").toURI()
).toFile()
);

private static byte[] serialize(Object o) throws Exception {
ByteArrayOutputStream bos = new ByteArrayOutputStream();
try (ObjectOutput out = new ObjectOutputStream(bos)) {
out.writeObject(o);
EasyPredictModelWrapper.Config config = new EasyPredictModelWrapper.Config()
.setModel(MojoModel.load(mojofile));
EasyPredictModelWrapper model = new EasyPredictModelWrapper(config);

String[][] inputs = new String[][]{
{"1.0", "5.0", "A"},
{"2.0", "4.0", "B"},
{"3.0", "3.0", "A"},
};

double [] offsets = {0.1, 0.2, 0.3};
double[] expected = {9.81646, 19.7366, 29.6913};
double[] preds = new double[inputs.length];
for (int i = 0; i < inputs.length; i++) {
RowData row = new RowData();
row.put("numeric1", inputs[i][0]);
row.put("numeric2", inputs[i][1]);
row.put("categorical", inputs[i][2]);
preds[i] = model.predictRegression(row, offsets[i]).value;
}

assertArrayEquals(expected, preds, 0.0001);
}
return bos.toByteArray();
}

private static Object deserialize(byte[] bs) throws Exception {
try (ByteArrayInputStream bis = new ByteArrayInputStream(bs)) {
ObjectInput in = new ObjectInputStream(bis);
return in.readObject();
/*
Test for XGBoost model trained with offset and predicting with zero offset.
*/
@Test
public void testXGBWithOffset_ZeroOffset() throws Exception {
String mojofile = String.valueOf(
Paths.get(
XGBoostJavaMojoModelTest.class.getClassLoader().getResource("hex/genmodel/algos/xgboost/XGBoostWithOffset.zip").toURI()
).toFile()
);

EasyPredictModelWrapper.Config config = new EasyPredictModelWrapper.Config()
.setModel(MojoModel.load(mojofile));
EasyPredictModelWrapper model = new EasyPredictModelWrapper(config);

String[][] inputs = new String[][]{
{"1.0", "5.0", "A"},
{"2.0", "4.0", "B"},
{"3.0", "3.0", "A"},
};

double[] offsets = {0.0, 0.0, 0.0};
double[] expected = {9.7164, 19.5366, 29.3913};
double[] preds = new double[inputs.length];
for (int i = 0; i < inputs.length; i++) {
RowData row = new RowData();
row.put("numeric1", inputs[i][0]);
row.put("numeric2", inputs[i][1]);
row.put("categorical", inputs[i][2]);
preds[i] = model.predictRegression(row, offsets[i]).value;
}

assertArrayEquals(expected, preds, 0.0001);
}
}

/*
Test for XGBoost model trained without offset and predicting with zero offset.
*/
@Test
public void testXGBWithoutOffset_ZeroOffset() throws Exception {
String mojofile = String.valueOf(
Paths.get(
XGBoostJavaMojoModelTest.class.getClassLoader().getResource("hex/genmodel/algos/xgboost/XGBoostWithoutOffset.zip").toURI()
).toFile()
);

EasyPredictModelWrapper.Config config = new EasyPredictModelWrapper.Config()
.setModel(MojoModel.load(mojofile));
EasyPredictModelWrapper model = new EasyPredictModelWrapper(config);

String[][] inputs = new String[][]{
{"1.0", "5.0", "A"},
{"2.0", "4.0", "B"},
{"3.0", "3.0", "A"},
};

double[] offsets = {0.0, 0.0, 0.0};
double[] expected = {9.8089, 19.7527, 29.6871};
double[] preds = new double[inputs.length];
for (int i = 0; i < inputs.length; i++) {
RowData row = new RowData();
row.put("numeric1", inputs[i][0]);
row.put("numeric2", inputs[i][1]);
row.put("categorical", inputs[i][2]);
preds[i] = model.predictRegression(row, offsets[i]).value;
}

assertArrayEquals(expected, preds, 0.0001);
}

private static byte[] serialize(Object o) throws Exception {
ByteArrayOutputStream bos = new ByteArrayOutputStream();
try (ObjectOutput out = new ObjectOutputStream(bos)) {
out.writeObject(o);
}
return bos.toByteArray();
}

private static Object deserialize(byte[] bs) throws Exception {
try (ByteArrayInputStream bis = new ByteArrayInputStream(bs)) {
ObjectInput in = new ObjectInputStream(bis);
return in.readObject();
}
}

}
Binary file added ...el-extensions/xgboost/src/test/resources/hex/genmodel/algos/xgboost/XGBoostWithOffset.zip
View file
Open in desktop
Binary file not shown.
Binary file added ...extensions/xgboost/src/test/resources/hex/genmodel/algos/xgboost/XGBoostWithoutOffset.zip
View file
Open in desktop
Binary file not shown.
Loading