Predict.

demo/dask/cpu_training.py

@@ -27,8 +27,7 @@ def main(client):
                              'nthread': 1,
                              'tree_method': 'hist'},
                             dtrain,
-                            num_boost_round=4, evals=[(dtrain, 'train')],
-                            host_ip='127.0.0.1', port=32809)
+                            num_boost_round=4, evals=[(dtrain, 'train')])
     bst = output['booster']
     history = output['history']
 

python-package/xgboost/dask.py

@@ -342,7 +342,8 @@ def get_worker_data_shape(self, worker):
 
 def _get_rabit_args(worker_map, client, host_ip=None, port=None):
     '''Get rabit context arguments from data distribution in DaskDMatrix.'''
-    assert (host_ip and port) or (host_ip is None and port is None)
+    msg = 'Please provide both IP and port'
+    assert (host_ip and port) or (host_ip is None and port is None), msg
 
     if host_ip:
         logger.info('Running tracker on: %s, %s', host_ip, str(port))
@@ -363,7 +364,7 @@ def _get_rabit_args(worker_map, client, host_ip=None, port=None):
 # evaluation history is instead returned.
 
 
-def train(client, params, dtrain, *args, evals=(), host_ip=None, port=None,
+def train(client, params, dtrain, *args, evals=(), tracker_ip=None, tracker_port=None,
           **kwargs):
     '''Train XGBoost model.
 
@@ -399,7 +400,7 @@ def train(client, params, dtrain, *args, evals=(), host_ip=None, port=None,
     client = _xgb_get_client(client)
     workers = list(_get_client_workers(client).keys())
 
-    rabit_args = _get_rabit_args(workers, client, host_ip, port)
+    rabit_args = _get_rabit_args(workers, client, tracker_ip, tracker_port)
 
     def dispatched_train(worker_addr):
         '''Perform training on a single worker.'''
@@ -438,7 +439,7 @@ def dispatched_train(worker_addr):
     return list(filter(lambda ret: ret is not None, results))[0]
 
 
-def predict(client, model, data, *args):
+def predict(client, model, data, *args, tracker_ip=None, tracker_port=None):
     '''Run prediction with a trained booster.
 
     .. note::
@@ -475,7 +476,7 @@ def predict(client, model, data, *args):
     worker_map = data.worker_map
     client = _xgb_get_client(client)
 
-    rabit_args = _get_rabit_args(worker_map, client)
+    rabit_args = _get_rabit_args(worker_map, client, tracker_ip, tracker_port)
 
     def dispatched_predict(worker_id):
         '''Perform prediction on each worker.'''