fix logger

README.md

@@ -169,6 +169,8 @@ And lastly, analyze the results.
 
 ```python
 atom.evaluate()
+
+atom.plot_lift()
 ```
 
 <br><br>

atom/api.py

@@ -272,7 +272,7 @@ class ATOMClassifier(BaseTransformer, ATOM):
         ATOM can't manage warnings that go from C/C++ code to stdout.
 
     logger: str, Logger or None, default=None
-        - If None: Doesn't save a logging file.
+        - If None: Logging isn't used.
         - If str: Name of the log file. Use "auto" for automatic name.
         - Else: Python `logging.Logger` instance.
 
@@ -498,7 +498,7 @@ class ATOMForecaster(BaseTransformer, ATOM):
         ATOM can't manage warnings that go from C/C++ code to stdout.
 
     logger: str, Logger or None, default=None
-        - If None: Doesn't save a logging file.
+        - If None: Logging isn't used.
         - If str: Name of the log file. Use "auto" for automatic name.
         - Else: Python `logging.Logger` instance.
 
@@ -727,7 +727,7 @@ class ATOMRegressor(BaseTransformer, ATOM):
         ATOM can't manage warnings that go from C/C++ code to stdout.
 
     logger: str, Logger or None, default=None
-        - If None: Doesn't save a logging file.
+        - If None: Logging isn't used.
         - If str: Name of the log file. Use "auto" for automatic name.
         - Else: Python `logging.Logger` instance.
 

atom/atom.py

@@ -19,7 +19,6 @@
 import numpy as np
 import pandas as pd
 from scipy import stats
-from sklearn.preprocessing import FunctionTransformer
 from sklearn.utils.metaestimators import available_if
 from sklearn.utils.validation import check_memory
 
@@ -45,12 +44,12 @@
     TrainSizingRegressor,
 )
 from atom.utils import (
-    DATAFRAME, FEATURES, INT, PANDAS, SCALAR, SEQUENCE, SERIES, TARGET,
-    TS_INDEX_TYPES, ClassMap, DataConfig, Predictor, Runner, Transformer,
-    __version__, bk, check_dependency, check_is_fitted, check_scaling,
+    DATAFRAME, FEATURES, INT, MISSING_VALUES, PANDAS, SCALAR, SEQUENCE, SERIES,
+    TARGET, TS_INDEX_TYPES, ClassMap, DataConfig, Predictor, Runner,
+    Transformer, __version__, check_dependency, check_is_fitted, check_scaling,
     composed, crash, custom_transform, fit_one, flt, get_cols,
     get_custom_scorer, has_task, infer_task, is_multioutput, is_sparse, lst,
-    method_to_log, sign, variable_return, MISSING_VALUES
+    method_to_log, sign, variable_return,
 )
 
 
@@ -846,7 +845,7 @@ def get_data(new_t: str) -> SERIES:
         }
 
         # Convert selected columns to the best nullable dtype
-        data = self.dataset[self.branch._get_columns(columns)]
+        data = self.dataset[self.branch._get_columns(columns)]  # TODO: .convert_dtypes()
 
         for name, column in data.items():
             if pd.api.types.is_sparse(column):
@@ -892,11 +891,11 @@ def get_data(new_t: str) -> SERIES:
         from pandas.core.dtypes.cast import convert_dtypes
         print(self.dtypes)
         self.branch.dataset = self.branch.dataset.astype(
-                {
-                    name: convert_dtypes(column, dtype_backend="pyarrow")
-                    for name, column in data.items()
-                }
-            )
+            {
+                name: convert_dtypes(column, dtype_backend="pyarrow")
+                for name, column in data.items()
+            }
+        )
 
         self.log("The column dtypes are successfully converted.", 1)
 
@@ -1264,8 +1263,10 @@ def apply(
             Additional keyword arguments for the inverse function.
 
         """
+        estimator = self._get_est_class("FunctionTransformer", "preprocessing")
+
         columns = kwargs.pop("columns", None)
-        function_transformer = FunctionTransformer(
+        function_transformer = estimator(
             func=func,
             inverse_func=inverse_func,
             kw_args=kw_args,

atom/basemodel.py

@@ -152,7 +152,7 @@ class BaseModel(BaseTransformer, BaseTracker, HTPlot, PredictionPlot, ShapPlot):
         ATOM can't manage warnings that go from C/C++ code to stdout.
 
     logger: str, Logger or None, default=None
-        - If None: Doesn't save a logging file.
+        - If None: Logging isn't used.
         - If str: Name of the log file. Use "auto" for automatic name.
         - Else: Python `logging.Logger` instance.
 

atom/baserunner.py

@@ -180,28 +180,32 @@ def metric(self) -> str | list[str] | None:
             return flt(self._metric.keys())
 
     @property
-    def winners(self) -> list[Model]:
+    def winners(self) -> list[Model] | None:
         """Models ordered by performance.
 
         Performance is measured as the highest score on the model's
         [`score_bootstrap`][adaboost-score_bootstrap] or
         [`score_test`][adaboost-score_test] attributes, checked in
         that order. For [multi-metric runs][], only the main metric
-        is compared.
+        is compared. Ties are resolved looking at the lowest
+        [time_fit][adaboost-time_fit].
 
         """
         if self._models:  # Returns None if not fitted
-            return sorted(self._models, key=lambda x: get_best_score(x), reverse=True)
+            return sorted(
+                self._models, key=lambda x: (get_best_score(x), x.time_fit), reverse=True
+            )
 
     @property
-    def winner(self) -> Model:
+    def winner(self) -> Model | None:
         """Best performing model.
 
         Performance is measured as the highest score on the model's
         [`score_bootstrap`][adaboost-score_bootstrap] or
         [`score_test`][adaboost-score_test] attributes, checked in
         that order. For [multi-metric runs][], only the main metric
-        is compared.
+        is compared. Ties are resolved looking at the lowest
+        [time_fit][adaboost-time_fit].
 
         """
         if self._models:  # Returns None if not fitted

atom/basetransformer.py

@@ -17,7 +17,7 @@
 from importlib import import_module
 from importlib.util import find_spec
 from logging import (
-    DEBUG, FileHandler, Formatter, Logger, NullHandler, getLogger,
+    DEBUG, FileHandler, Formatter, Logger, getLogger,
 )
 from multiprocessing import cpu_count
 from typing import Any, Callable
@@ -162,8 +162,11 @@ def engine(self, value: dict | None):
                 if not find_spec("cuml"):
                     raise ModuleNotFoundError(
                         "Failed to import cuml. Package is not installed. Refer "
-                        "to: https://rapids.ai/start.html#rapids-release-selector."
+                        "to: https://rapids.ai/start.html#install."
                     )
+                else:
+                    import cuml
+                    cuml.internals.memory_utils.set_global_output_type("numpy")
             elif models.lower() != "sklearn":
                 raise ValueError(
                     "Invalid value for the models key of the engine parameter, "
@@ -234,8 +237,8 @@ def warnings(self, value: bool | str):
         os.environ["PYTHONWARNINGS"] = self._warnings  # Affects subprocesses (joblib)
 
     @property
-    def logger(self) -> Logger:
-        """Logger of this class."""
+    def logger(self) -> Logger | None:
+        """Logger for this instance."""
         return self._logger
 
     @logger.setter
@@ -250,37 +253,39 @@ def logger(self, value: str | Logger | None):
             "evalml",
         ]
 
-        if isinstance(value, Logger):
-            logger = value
+        # Clear existing handlers for external loggers
+        for name in external_loggers:
+            for handler in (log := getLogger(name)).handlers:
+                handler.close()
+            log.handlers.clear()
+
+        if not value:
+            logger = None
         else:
-            logger = getLogger(self.__class__.__name__)
-            logger.setLevel(DEBUG)
+            if isinstance(value, Logger):
+                logger = value
+            else:
+                logger = getLogger(self.__class__.__name__)
+                logger.setLevel(DEBUG)
 
-            # Clear existing handlers
-            for name in [logger.name] + external_loggers:
-                for handler in (log := getLogger(name)).handlers:
+                # Clear existing handlers for current logger
+                for handler in logger.handlers:
                     handler.close()
-                log.handlers.clear()
+                logger.handlers.clear()
 
-            # Setup file handler
-            if isinstance(value, str):
                 # Prepare the FileHandler
                 if not value.endswith(".log"):
                     value += ".log"
                 if os.path.basename(value) == "auto.log":
                     current = dt.now().strftime("%d%b%y_%Hh%Mm%Ss")
                     value = value.replace("auto", self.__class__.__name__ + "_" + current)
 
-                formatter = Formatter("%(asctime)s - %(levelname)s: %(message)s")
+                fh = FileHandler(value)
+                fh.setFormatter(Formatter("%(asctime)s - %(levelname)s: %(message)s"))
 
-                handler = FileHandler(value)
-                handler.setFormatter(formatter)
-            else:
-                handler = NullHandler()
-
-            # Redirect loggers to file handler
-            for log in [logger.name] + external_loggers:
-                getLogger(log).addHandler(handler)
+                # Redirect loggers to file handler
+                for log in [logger.name] + external_loggers:
+                    getLogger(log).addHandler(fh)
 
         self._logger = logger
 
@@ -751,6 +756,8 @@ def _no_data_sets(
             else:
                 test_size = self.test_size
 
+            splitter = self._get_est_class("train_test_split", "model_selection")
+
             try:
                 # Define holdout set size
                 if self.holdout_size:
@@ -766,7 +773,7 @@ def _no_data_sets(
                             f"got {self.holdout_size}."
                         )
 
-                    data, holdout = train_test_split(
+                    data, holdout = splitter(
                         data,
                         test_size=holdout_size,
                         random_state=self.random_state,
@@ -777,7 +784,7 @@ def _no_data_sets(
                 else:
                     holdout = None
 
-                train, test = train_test_split(
+                train, test = splitter(
                     data,
                     test_size=test_size,
                     random_state=self.random_state,
@@ -1018,8 +1025,9 @@ def log(self, msg: SCALAR | str, level: INT = 0, severity: str = "info"):
         elif severity == "info" and self.verbose >= level:
             print(msg)
 
-        for text in str(msg).split("\n"):
-            getattr(self.logger, severity)(str(text))
+        if self.logger:
+            for text in str(msg).split("\n"):
+                getattr(self.logger, severity)(str(text))
 
     @composed(crash, method_to_log)
     def save(self, filename: str = "auto", *, save_data: bool = True):

atom/data_cleaning.py

@@ -41,7 +41,6 @@
 from scipy.stats import zscore
 from sklearn.base import BaseEstimator, clone
 from sklearn.impute import KNNImputer
-from sklearn.preprocessing import FunctionTransformer
 
 from atom.basetransformer import BaseTransformer
 from atom.utils import (
@@ -226,7 +225,7 @@ class Balancer(BaseEstimator, TransformerMixin, BaseTransformer):
         - 2 to print detailed information.
 
     logger: str, Logger or None, default=None
-        - If None: Doesn't save a logging file.
+        - If None: Logging isn't used.
         - If str: Name of the log file. Use "auto" for automatic naming.
         - Else: Python `logging.Logger` instance.
 
@@ -571,7 +570,7 @@ class Cleaner(BaseEstimator, TransformerMixin, BaseTransformer):
         - 2 to print detailed information.
 
     logger: str, Logger or None, default=None
-        - If None: Doesn't save a logging file.
+        - If None: Logging isn't used.
         - If str: Name of the log file. Use "auto" for automatic naming.
         - Else: Python `logging.Logger` instance.
 
@@ -1004,7 +1003,7 @@ class Discretizer(BaseEstimator, TransformerMixin, BaseTransformer):
         - 2 to print detailed information.
 
     logger: str, Logger or None, default=None
-        - If None: Doesn't save a logging file.
+        - If None: Logging isn't used.
         - If str: Name of the log file. Use "auto" for automatic naming.
         - Else: Python `logging.Logger` instance.
 
@@ -1207,8 +1206,10 @@ def get_labels(labels, bins):
                 else:
                     bins = [-np.inf] + list(bins) + [np.inf]
 
+                estimator = self._get_est_class("FunctionTransformer", "preprocessing")
+
                 # Make of cut a transformer
-                self._discretizers[col] = FunctionTransformer(
+                self._discretizers[col] = estimator(
                     func=bk.cut,
                     kw_args={"bins": bins, "labels": get_labels(labels, bins)},
                 ).fit(X[[col]])
@@ -1316,7 +1317,7 @@ class Encoder(BaseEstimator, TransformerMixin, BaseTransformer):
         - 2 to print detailed information.
 
     logger: str, Logger or None, default=None
-        - If None: Doesn't save a logging file.
+        - If None: Logging isn't used.
         - If str: Name of the log file. Use "auto" for automatic naming.
         - Else: Python `logging.Logger` instance.
 
@@ -1724,7 +1725,7 @@ class Imputer(BaseEstimator, TransformerMixin, BaseTransformer):
         - 2 to print detailed information.
 
     logger: str, Logger or None, default=None
-        - If None: Doesn't save a logging file.
+        - If None: Logging isn't used.
         - If str: Name of the log file. Use "auto" for automatic naming.
         - Else: Python `logging.Logger` instance.
 
@@ -2124,7 +2125,7 @@ class Normalizer(BaseEstimator, TransformerMixin, BaseTransformer):
         - 1 to print basic information.
 
     logger: str, Logger or None, default=None
-        - If None: Doesn't save a logging file.
+        - If None: Logging isn't used.
         - If str: Name of the log file. Use "auto" for automatic naming.
         - Else: Python `logging.Logger` instance.
 
@@ -2425,7 +2426,7 @@ class Pruner(BaseEstimator, TransformerMixin, BaseTransformer):
         - 2 to print detailed information.
 
     logger: str, Logger or None, default=None
-        - If None: Doesn't save a logging file.
+        - If None: Logging isn't used.
         - If str: Name of the log file. Use "auto" for automatic naming.
         - Else: Python `logging.Logger` instance.
 
@@ -2728,7 +2729,7 @@ class Scaler(BaseEstimator, TransformerMixin, BaseTransformer):
         - 1 to print basic information.
 
     logger: str, Logger or None, default=None
-        - If None: Doesn't save a logging file.
+        - If None: Logging isn't used.
         - If str: Name of the log file. Use "auto" for automatic naming.
         - Else: Python `logging.Logger` instance.