Fix CLI model IO. (#5535)

* Add test for comparing Python and CLI training result.

Jenkinsfile

@@ -201,6 +201,7 @@ def BuildCPU() {
     ${docker_extra_params} ${dockerRun} ${container_type} ${docker_binary} build/testxgboost
     """
 
+    stash name: 'xgboost_cli', includes: 'xgboost'
     deleteDir()
   }
 }
@@ -282,6 +283,7 @@ def TestPythonCPU() {
   node('linux && cpu') {
     unstash name: 'xgboost_whl_cuda9'
     unstash name: 'srcs'
+    unstash name: 'xgboost_cli'
     echo "Test Python CPU"
     def container_type = "cpu"
     def docker_binary = "docker"

Jenkinsfile-win64

@@ -96,6 +96,7 @@ def BuildWin64() {
     s3Upload bucket: 'xgboost-nightly-builds', path: path, acl: 'PublicRead', workingDir: 'python-package/dist', includePathPattern:'**/*.whl'
     echo 'Stashing C++ test executable (testxgboost)...'
     stash name: 'xgboost_cpp_tests', includes: 'build/testxgboost.exe'
+    stash name: 'xgboost_cli', includes: 'xgboost.exe'
     deleteDir()
   }
 }
@@ -104,6 +105,7 @@ def TestWin64CPU() {
   node('win64 && cpu') {
     unstash name: 'srcs'
     unstash name: 'xgboost_whl'
+    unstash name: 'xgboost_cli'
     echo "Test Win64 CPU"
     echo "Installing Python wheel..."
     bat "conda activate && (python -m pip uninstall -y xgboost || cd .)"

src/cli_main.cc

@@ -138,14 +138,10 @@ struct CLIParam : public XGBoostParameter<CLIParam> {
     // constraint.
     if (name_pred == "stdout") {
       save_period = 0;
-      this->cfg.emplace_back(std::make_pair("silent", "0"));
     }
     if (dsplit == 0 && rabit::IsDistributed()) {
       dsplit = 2;
     }
-    if (rabit::GetRank() != 0) {
-      this->cfg.emplace_back(std::make_pair("silent", "1"));
-    }
   }
 };
 
@@ -189,7 +185,7 @@ void CLITrain(const CLIParam& param) {
     if (param.model_in != "NULL") {
       std::unique_ptr<dmlc::Stream> fi(
           dmlc::Stream::Create(param.model_in.c_str(), "r"));
-      learner->Load(fi.get());
+      learner->LoadModel(fi.get());
       learner->SetParams(param.cfg);
     } else {
       learner->SetParams(param.cfg);
@@ -229,7 +225,7 @@ void CLITrain(const CLIParam& param) {
          << i + 1 << ".model";
       std::unique_ptr<dmlc::Stream> fo(
           dmlc::Stream::Create(os.str().c_str(), "w"));
-      learner->Save(fo.get());
+      learner->SaveModel(fo.get());
     }
 
     if (learner->AllowLazyCheckPoint()) {
@@ -255,7 +251,7 @@ void CLITrain(const CLIParam& param) {
     }
     std::unique_ptr<dmlc::Stream> fo(
         dmlc::Stream::Create(os.str().c_str(), "w"));
-    learner->Save(fo.get());
+    learner->SaveModel(fo.get());
   }
 
   double elapsed = dmlc::GetTime() - start;
@@ -277,7 +273,7 @@ void CLIDumpModel(const CLIParam& param) {
   std::unique_ptr<dmlc::Stream> fi(
       dmlc::Stream::Create(param.model_in.c_str(), "r"));
   learner->SetParams(param.cfg);
-  learner->Load(fi.get());
+  learner->LoadModel(fi.get());
   // dump data
   std::vector<std::string> dump = learner->DumpModel(
       fmap, param.dump_stats, param.dump_format);
@@ -316,7 +312,7 @@ void CLIPredict(const CLIParam& param) {
   std::unique_ptr<Learner> learner(Learner::Create({}));
   std::unique_ptr<dmlc::Stream> fi(
       dmlc::Stream::Create(param.model_in.c_str(), "r"));
-  learner->Load(fi.get());
+  learner->LoadModel(fi.get());
   learner->SetParams(param.cfg);
 
   LOG(INFO) << "start prediction...";

tests/python/test_cli.py

@@ -0,0 +1,91 @@
+import os
+import tempfile
+import unittest
+import platform
+import xgboost
+import subprocess
+import numpy
+
+
+class TestCLI(unittest.TestCase):
+    template = '''
+booster = gbtree
+objective = reg:squarederror
+eta = 1.0
+gamma = 1.0
+seed = 0
+min_child_weight = 0
+max_depth = 3
+task = {task}
+model_in = {model_in}
+model_out = {model_out}
+test_path = {test_path}
+name_pred = {name_pred}
+
+num_round = 10
+data = {data_path}
+eval[test] = {data_path}
+'''
+
+    def test_cli_model(self):
+        curdir = os.path.normpath(os.path.abspath(os.path.dirname(__file__)))
+        project_root = os.path.normpath(
+            os.path.join(curdir, os.path.pardir, os.path.pardir))
+        data_path = "{root}/demo/data/agaricus.txt.train?format=libsvm".format(
+            root=project_root)
+
+        if platform.system() == 'Windows':
+            exe = 'xgboost.exe'
+        else:
+            exe = 'xgboost'
+        exe = os.path.join(project_root, exe)
+        assert os.path.exists(exe)
+
+        with tempfile.TemporaryDirectory() as tmpdir:
+            model_out = os.path.join(tmpdir, 'test_load_cli_model')
+            config_path = os.path.join(tmpdir, 'test_load_cli_model.conf')
+
+            train_conf = self.template.format(data_path=data_path,
+                                              task='train',
+                                              model_in='NULL',
+                                              model_out=model_out,
+                                              test_path='NULL',
+                                              name_pred='NULL')
+            with open(config_path, 'w') as fd:
+                fd.write(train_conf)
+
+            subprocess.run([exe, config_path])
+
+            predict_out = os.path.join(tmpdir,
+                                       'test_load_cli_model-prediction')
+            predict_conf = self.template.format(task='pred',
+                                                data_path=data_path,
+                                                model_in=model_out,
+                                                model_out='NULL',
+                                                test_path=data_path,
+                                                name_pred=predict_out)
+            with open(config_path, 'w') as fd:
+                fd.write(predict_conf)
+
+            subprocess.run([exe, config_path])
+
+            cli_predt = numpy.loadtxt(predict_out)
+
+            parameters = {
+                'booster': 'gbtree',
+                'objective': 'reg:squarederror',
+                'eta': 1.0,
+                'gamma': 1.0,
+                'seed': 0,
+                'min_child_weight': 0,
+                'max_depth': 3
+            }
+            data = xgboost.DMatrix(data_path)
+            booster = xgboost.train(parameters, data, num_boost_round=10)
+            py_predt = booster.predict(data)
+
+            numpy.testing.assert_allclose(cli_predt, py_predt)
+
+            cli_model = xgboost.Booster(model_file=model_out)
+            cli_predt = cli_model.predict(data)
+            numpy.testing.assert_allclose(cli_predt, py_predt)