[python] Dataset params back up before training (#786)

* params back up

* refine logic

python-package/lightgbm/basic.py

@@ -4,6 +4,7 @@
 """Wrapper c_api of LightGBM"""
 from __future__ import absolute_import
 
+import copy
 import ctypes
 import os
 import warnings
@@ -591,11 +592,12 @@ class Dataset(object):
         self.silent = silent
         self.feature_name = feature_name
         self.categorical_feature = categorical_feature
-        self.params = params
+        self.params = copy.deepcopy(params)
         self.free_raw_data = free_raw_data
         self.used_indices = None
         self._predictor = None
         self.pandas_categorical = None
+        self.params_back_up = None
 
     def __del__(self):
         self._free_handle()
@@ -872,8 +874,13 @@ class Dataset(object):
         if not self.params:
             self.params = params
         else:
+            self.params_back_up = copy.deepcopy(self.params)
             self.params.update(params)
 
+    def _reverse_update_params(self):
+        self.params = copy.deepcopy(self.params_back_up)
+        self.params_back_up = None
+
     def set_field(self, field_name, data):
         """Set property into the Dataset.
 

python-package/lightgbm/engine.py

@@ -128,14 +128,13 @@ def train(params, train_set, num_boost_round=100,
                 continue
             if not isinstance(valid_data, Dataset):
                 raise TypeError("Traninig only accepts Dataset object")
+            valid_data._update_params(params)
             valid_data.set_reference(train_set)
             reduced_valid_sets.append(valid_data)
             if valid_names is not None and len(valid_names) > i:
                 name_valid_sets.append(valid_names[i])
             else:
                 name_valid_sets.append('valid_' + str(i))
-        for valid_data in valid_sets:
-            valid_data._update_params(params)
     """process callbacks"""
     if callbacks is None:
         callbacks = set()
@@ -165,11 +164,16 @@ def train(params, train_set, num_boost_round=100,
     callbacks_after_iter = sorted(callbacks_after_iter, key=attrgetter('order'))
 
     """construct booster"""
+    try:
         booster = Booster(params=params, train_set=train_set)
         if is_valid_contain_train:
             booster.set_train_data_name(train_data_name)
         for valid_set, name_valid_set in zip(reduced_valid_sets, name_valid_sets):
             booster.add_valid(valid_set, name_valid_set)
+    finally:
+        train_set._reverse_update_params()
+        for valid_set in reduced_valid_sets:
+            valid_set._reverse_update_params()
     booster.best_iteration = 0
 
     """start training"""