diff --git a/python-package/xgboost/training.py b/python-package/xgboost/training.py
index 74223a166..7a29d3952 100644
--- a/python-package/xgboost/training.py
+++ b/python-package/xgboost/training.py
@@ -17,6 +17,7 @@ def _train_internal(params, dtrain,
     """internal training function"""
     callbacks = [] if callbacks is None else callbacks
     evals = list(evals)
+    params = params.copy()
     if isinstance(params, dict) \
             and 'eval_metric' in params \
             and isinstance(params['eval_metric'], list):
diff --git a/src/cli_main.cc b/src/cli_main.cc
index 3326b8cd5..f9f3b59e4 100644
--- a/src/cli_main.cc
+++ b/src/cli_main.cc
@@ -340,7 +340,6 @@ int CLIRunTask(int argc, char *argv[]) {
 
   common::ConfigParser cp(argv[1]);
   auto cfg = cp.Parse();
-  cfg.emplace_back("seed", "0");
 
   for (int i = 2; i < argc; ++i) {
     char name[256], val[256];
diff --git a/tests/python/test_cli.py b/tests/python/test_cli.py
index f1d1a863f..b3391b7b1 100644
--- a/tests/python/test_cli.py
+++ b/tests/python/test_cli.py
@@ -13,7 +13,7 @@ booster = gbtree
 objective = reg:squarederror
 eta = 1.0
 gamma = 1.0
-seed = 0
+seed = {seed}
 min_child_weight = 0
 max_depth = 3
 task = {task}
@@ -41,14 +41,18 @@ eval[test] = {data_path}
         exe = os.path.join(project_root, exe)
         assert os.path.exists(exe)
 
+        seed = 1994
+
         with tempfile.TemporaryDirectory() as tmpdir:
-            model_out = os.path.join(tmpdir, 'test_load_cli_model')
+            model_out_cli = os.path.join(tmpdir, 'test_load_cli_model-cli.bin')
+            model_out_py = os.path.join(tmpdir, 'test_cli_model-py.bin')
             config_path = os.path.join(tmpdir, 'test_load_cli_model.conf')
 
             train_conf = self.template.format(data_path=data_path,
+                                              seed=seed,
                                               task='train',
                                               model_in='NULL',
-                                              model_out=model_out,
+                                              model_out=model_out_cli,
                                               test_path='NULL',
                                               name_pred='NULL')
             with open(config_path, 'w') as fd:
@@ -59,8 +63,9 @@ eval[test] = {data_path}
             predict_out = os.path.join(tmpdir,
                                        'test_load_cli_model-prediction')
             predict_conf = self.template.format(task='pred',
+                                                seed=seed,
                                                 data_path=data_path,
-                                                model_in=model_out,
+                                                model_in=model_out_cli,
                                                 model_out='NULL',
                                                 test_path=data_path,
                                                 name_pred=predict_out)
@@ -76,16 +81,24 @@ eval[test] = {data_path}
                 'objective': 'reg:squarederror',
                 'eta': 1.0,
                 'gamma': 1.0,
-                'seed': 0,
+                'seed': seed,
                 'min_child_weight': 0,
                 'max_depth': 3
             }
             data = xgboost.DMatrix(data_path)
             booster = xgboost.train(parameters, data, num_boost_round=10)
+            booster.save_model(model_out_py)
             py_predt = booster.predict(data)
 
             numpy.testing.assert_allclose(cli_predt, py_predt)
 
-            cli_model = xgboost.Booster(model_file=model_out)
+            cli_model = xgboost.Booster(model_file=model_out_cli)
             cli_predt = cli_model.predict(data)
             numpy.testing.assert_allclose(cli_predt, py_predt)
+
+            with open(model_out_cli, 'rb') as fd:
+                cli_model_bin = fd.read()
+            with open(model_out_py, 'rb') as fd:
+                py_model_bin = fd.read()
+
+            assert hash(cli_model_bin) == hash(py_model_bin)