[python] pass params to _InnerPredictor in train and cv and verbose f…

…ix (#1628)

* pass params to _InnerPredictor in train and cv

* fixed verbosity param description

* treat silent param as Fatal log level

* create Dataset in refit method silently

* do not overwrite verbose param by silent argument

docs/Parameters.rst

@@ -377,7 +377,7 @@ IO Parameters
 
    -  controls the level of LightGBM's verbosity
 
-   -  ``< 0``: Fatal, ``= 0``: Error (Warn), ``> 0``: Info
+   -  ``< 0``: Fatal, ``= 0``: Error (Warning), ``= 1``: Info, ``> 1``: Debug
 
 -  ``max_bin`` :raw-html:`<a id="max_bin" title="Permalink to this parameter" href="#max_bin">&#x1F517;&#xFE0E;</a>`, default = ``255``, type = int, constraints: ``max_bin > 1``
 

include/LightGBM/config.h

@@ -380,7 +380,7 @@ struct Config {
 
   // alias = verbose
   // desc = controls the level of LightGBM's verbosity
-  // desc = ``< 0``: Fatal, ``= 0``: Error (Warn), ``> 0``: Info
+  // desc = ``< 0``: Fatal, ``= 0``: Error (Warning), ``= 1``: Info, ``> 1``: Debug
   int verbosity = 1;
 
   // check = >1

python-package/lightgbm/basic.py

@@ -701,8 +701,9 @@ def _lazy_init(self, data, label=None, reference=None,
                 warnings.warn('{0} keyword has been found in `params` and will be ignored. '
                               'Please use {0} argument of the Dataset constructor to pass this parameter.'.format(key))
         self.predictor = predictor
-        if silent:
-            params["verbose"] = 0
+        # user can set verbose with params, it has higher priority
+        if not any(verbose_alias in params for verbose_alias in ('verbose', 'verbosity')) and silent:
+            params["verbose"] = -1
         # get categorical features
         if categorical_feature is not None:
             categorical_indices = set()
@@ -1431,8 +1432,9 @@ def __init__(self, params=None, train_set=None, model_file=None, silent=False):
         self.best_iteration = -1
         self.best_score = {}
         params = {} if params is None else params
-        if silent:
-            params["verbose"] = 0
+        # user can set verbose with params, it has higher priority
+        if not any(verbose_alias in params for verbose_alias in ('verbose', 'verbosity')) and silent:
+            params["verbose"] = -1
         if train_set is not None:
             # Training task
             if not isinstance(train_set, Dataset):
@@ -2118,7 +2120,7 @@ def refit(self, data, label, decay_rate=0.9, **kwargs):
         predictor = self._to_predictor()
         leaf_preds = predictor.predict(data, -1, pred_leaf=True, **kwargs)
         nrow, ncol = leaf_preds.shape
-        train_set = Dataset(data, label)
+        train_set = Dataset(data, label, silent=True)
         new_booster = Booster(self.params, train_set, silent=True)
         # Copy models
         _safe_call(_LIB.LGBM_BoosterMerge(

python-package/lightgbm/engine.py

@@ -118,9 +118,9 @@ def train(params, train_set, num_boost_round=100,
     if num_boost_round <= 0:
         raise ValueError("num_boost_round should be greater than zero.")
     if isinstance(init_model, string_type):
-        predictor = _InnerPredictor(model_file=init_model)
+        predictor = _InnerPredictor(model_file=init_model, pred_parameter=params)
     elif isinstance(init_model, Booster):
-        predictor = init_model._to_predictor()
+        predictor = init_model._to_predictor(dict(init_model.params, **params))
     else:
         predictor = None
     init_iteration = predictor.num_total_iteration if predictor is not None else 0
@@ -415,9 +415,9 @@ def cv(params, train_set, num_boost_round=100,
     if num_boost_round <= 0:
         raise ValueError("num_boost_round should be greater than zero.")
     if isinstance(init_model, string_type):
-        predictor = _InnerPredictor(model_file=init_model)
+        predictor = _InnerPredictor(model_file=init_model, pred_parameter=params)
     elif isinstance(init_model, Booster):
-        predictor = init_model._to_predictor()
+        predictor = init_model._to_predictor(dict(init_model.params, **params))
     else:
         predictor = None
     train_set._update_params(params)._set_predictor(predictor).set_feature_name(feature_name).set_categorical_feature(categorical_feature)

python-package/lightgbm/sklearn.py

@@ -402,7 +402,7 @@ def fit(self, X, y,
         params = self.get_params()
         # user can set verbose with kwargs, it has higher priority
         if not any(verbose_alias in params for verbose_alias in ('verbose', 'verbosity')) and self.silent:
-            params['verbose'] = 0
+            params['verbose'] = -1
         params.pop('silent', None)
         params.pop('importance_type', None)
         params.pop('n_estimators', None)