diff --git a/python/xorbits/_mars/learn/contrib/xgboost/tracker.py b/python/xorbits/_mars/learn/contrib/xgboost/tracker.py
index aeaca07cb..b2f02c078 100644
--- a/python/xorbits/_mars/learn/contrib/xgboost/tracker.py
+++ b/python/xorbits/_mars/learn/contrib/xgboost/tracker.py
@@ -241,7 +241,7 @@ def worker_envs(self) -> Dict[str, Union[str, int]]:
         get environment variables for workers
         can be passed in as args or envs
         """
-        return {"DMLC_TRACKER_URI": self.host_ip, "DMLC_TRACKER_PORT": self.port}
+        return {"rabbit_tracker_uri": self.host_ip, "rabbit_tracker_port": self.port}
 
     def _get_tree(self, n_workers: int) -> Tuple[_TreeMap, Dict[int, int]]:
         tree_map: _TreeMap = {}
diff --git a/python/xorbits/_mars/learn/contrib/xgboost/train.py b/python/xorbits/_mars/learn/contrib/xgboost/train.py
index 6ea49de4d..5ca4476db 100644
--- a/python/xorbits/_mars/learn/contrib/xgboost/train.py
+++ b/python/xorbits/_mars/learn/contrib/xgboost/train.py
@@ -224,18 +224,11 @@ def execute(cls, ctx, op: "XGBTrain"):
                 for arg in rabit_args
             ]
             parsed = {}
-            args_map_dmlc2rabit = {
-                "DMLC_TRACKER_URI": "rabit_tracker_uri",
-                "DMLC_TRACKER_PORT": "rabit_tracker_port",
-                "DMLC_NUM_WORKER": "rabit_num_worker",
-            }
             if rabit_args:
                 for arg in rabit_args:
                     kv = arg.decode().split("=")
                     if len(kv) == 2:
-                        parsed[args_map_dmlc2rabit[kv[0]]] = (
-                            int(kv[1]) if kv[0] == "DMLC_TRACKER_PORT" else kv[1]
-                        )
+                        parsed[kv[0]] = kv[1]
             collective.init(**parsed)
             try:
                 logger.debug(
diff --git a/test.py b/test.py
new file mode 100644
index 000000000..85d49e5c2
--- /dev/null
+++ b/test.py
@@ -0,0 +1,57 @@
+import pytest
+from python.xorbits import numpy as np
+from python.xorbits import pandas as pd
+from python.xorbits import xgboost as xxgb
+from python.xorbits._mars.core.entity.objects import ObjectData
+
+
+X = np.random.rand(100, 10)
+X_df = pd.DataFrame(X)
+y = np.random.randint(0, 2, 100)
+
+classifier = xxgb.XGBClassifier(verbosity=1, n_estimators=2)
+
+classifier.fit(X_df, y, eval_set=[(X_df, y)])
+pred = classifier.predict(X_df)
+
+assert pred.ndim == 1
+assert pred.shape[0] == len(X_df)
+
+history = classifier.evals_result()
+
+assert isinstance(history, dict)
+
+assert list(history)[0] == "validation_0"
+
+prob = classifier.predict_proba(X_df)
+
+assert prob.shape[0] == X_df.shape[0]
+
+assert len(pred) == len(y)
+assert set(pred.to_numpy().to_numpy()).issubset({0, 1})
+
+# test weight
+weights = [
+    np.random.rand(X_df.shape[0]),
+    pd.Series(np.random.rand(X_df.shape[0])),
+    pd.DataFrame(np.random.rand(X_df.shape[0])),
+]
+y_df = pd.DataFrame(y)
+for weight in weights:
+    classifier.fit(X_df, y_df, sample_weight=weight)
+    prediction = classifier.predict(X_df)
+
+    assert prediction.ndim == 1
+    assert prediction.shape[0] == len(X_df)
+
+# should raise error if weight.ndim > 1
+with pytest.raises(ValueError):
+    classifier.fit(X_df, y_df, sample_weight=np.random.rand(1, 1))
+
+# test wrong argument
+with pytest.raises(TypeError):
+    classifier.fit(X_df, y, wrong_param=1)
+
+# test wrong attribute
+with pytest.raises(AttributeError):
+    classifier.wrong_attribute()