[Logging] Reduce numpy 1.20 deprecation warnings from scipy (open-mml…

…ab#910)

core/src/autogluon/core/models/abstract/abstract_model.py

@@ -5,6 +5,7 @@
 import pickle
 import sys
 import time
+import warnings
 from typing import Union
 
 import numpy as np
@@ -299,12 +300,17 @@ def _preprocess_fit_resources(self, silent=False, **kwargs):
         return kwargs
 
     def fit(self, **kwargs):
-        kwargs = self._preprocess_fit_args(**kwargs)
-        if 'time_limit' not in kwargs or kwargs['time_limit'] is None or kwargs['time_limit'] > 0:
-            self._fit(**kwargs)
-        else:
-            logger.warning(f'\tWarning: Model has no time left to train, skipping model... (Time Left = {round(kwargs["time_limit"], 1)}s)')
-            raise TimeLimitExceeded
+        with warnings.catch_warnings():
+            # Suppress numpy 1.20 warnings (downstream scipy is not updated yet)
+            # https://numpy.org/devdocs/release/1.20.0-notes.html#deprecations
+            warnings.filterwarnings('ignore', message='`np.*` is a deprecated alias for the builtin `.*`')
+
+            kwargs = self._preprocess_fit_args(**kwargs)
+            if 'time_limit' not in kwargs or kwargs['time_limit'] is None or kwargs['time_limit'] > 0:
+                self._fit(**kwargs)
+            else:
+                logger.warning(f'\tWarning: Model has no time left to train, skipping model... (Time Left = {round(kwargs["time_limit"], 1)}s)')
+                raise TimeLimitExceeded
 
     def _fit(self, X_train, y_train, **kwargs):
         # kwargs may contain: num_cpus, num_gpus
@@ -326,23 +332,28 @@ def predict_proba(self, X, normalize=None, **kwargs):
         return y_pred_proba
 
     def _predict_proba(self, X, **kwargs):
-        X = self.preprocess(X, **kwargs)
-
-        if self.problem_type == REGRESSION:
-            return self.model.predict(X)
-
-        y_pred_proba = self.model.predict_proba(X)
-        if self.problem_type == BINARY:
-            if len(y_pred_proba.shape) == 1:
+        with warnings.catch_warnings():
+            # Suppress numpy 1.20 warnings (downstream scipy is not updated yet)
+            # https://numpy.org/devdocs/release/1.20.0-notes.html#deprecations
+            warnings.filterwarnings('ignore', message='`np.*` is a deprecated alias for the builtin `.*`')
+
+            X = self.preprocess(X, **kwargs)
+
+            if self.problem_type == REGRESSION:
+                return self.model.predict(X)
+
+            y_pred_proba = self.model.predict_proba(X)
+            if self.problem_type == BINARY:
+                if len(y_pred_proba.shape) == 1:
+                    return y_pred_proba
+                elif y_pred_proba.shape[1] > 1:
+                    return y_pred_proba[:, 1]
+                else:
+                    return y_pred_proba
+            elif y_pred_proba.shape[1] > 2:
                 return y_pred_proba
-            elif y_pred_proba.shape[1] > 1:
-                return y_pred_proba[:, 1]
             else:
-                return y_pred_proba
-        elif y_pred_proba.shape[1] > 2:
-            return y_pred_proba
-        else:
-            return y_pred_proba[:, 1]
+                return y_pred_proba[:, 1]
 
     def score(self, X, y, eval_metric=None, metric_needs_y_pred=None, **kwargs):
         if eval_metric is None:

tabular/src/autogluon/tabular/trainer/auto_trainer.py

@@ -1,4 +1,6 @@
 import logging
+import warnings
+
 import pandas as pd
 
 from autogluon.core.utils import generate_train_test_split
@@ -26,25 +28,30 @@ def get_models(self, hyperparameters, **kwargs):
                                  feature_metadata=feature_metadata, silent=silent, **kwargs)
 
     def fit(self, X_train, y_train, hyperparameters, X_val=None, y_val=None, X_unlabeled=None, feature_prune=False, holdout_frac=0.1, num_stack_levels=0, core_kwargs: dict = None, time_limit=None, **kwargs):
-        for key in kwargs:
-            logger.warning(f'Warning: Unknown argument passed to `AutoTrainer.fit()`. Argument: {key}')
+        with warnings.catch_warnings():
+            # Suppress numpy 1.20 warnings (downstream scipy is not updated yet)
+            # https://numpy.org/devdocs/release/1.20.0-notes.html#deprecations
+            warnings.filterwarnings('ignore', message='`np.*` is a deprecated alias for the builtin `.*`')
+
+            for key in kwargs:
+                logger.warning(f'Warning: Unknown argument passed to `AutoTrainer.fit()`. Argument: {key}')
 
-        if self.bagged_mode:
-            if (y_val is not None) and (X_val is not None):
-                # TODO: User could be intending to blend instead. Perhaps switch from OOF preds to X_val preds while still bagging? Doubt a user would want this.
-                logger.warning('Warning: Training AutoGluon in Bagged Mode but X_val is specified, concatenating X_train and X_val for cross-validation')
-                X_train = pd.concat([X_train, X_val], ignore_index=True)
-                y_train = pd.concat([y_train, y_val], ignore_index=True)
-            X_val = None
-            y_val = None
-        else:
-            if (y_val is None) or (X_val is None):
-                X_train, X_val, y_train, y_val = generate_train_test_split(X_train, y_train, problem_type=self.problem_type, test_size=holdout_frac, random_state=self.random_state)
-                logger.log(20, f'Automatically generating train/validation split with holdout_frac={holdout_frac}, Train Rows: {len(X_train)}, Val Rows: {len(X_val)}')
+            if self.bagged_mode:
+                if (y_val is not None) and (X_val is not None):
+                    # TODO: User could be intending to blend instead. Perhaps switch from OOF preds to X_val preds while still bagging? Doubt a user would want this.
+                    logger.warning('Warning: Training AutoGluon in Bagged Mode but X_val is specified, concatenating X_train and X_val for cross-validation')
+                    X_train = pd.concat([X_train, X_val], ignore_index=True)
+                    y_train = pd.concat([y_train, y_val], ignore_index=True)
+                X_val = None
+                y_val = None
+            else:
+                if (y_val is None) or (X_val is None):
+                    X_train, X_val, y_train, y_val = generate_train_test_split(X_train, y_train, problem_type=self.problem_type, test_size=holdout_frac, random_state=self.random_state)
+                    logger.log(20, f'Automatically generating train/validation split with holdout_frac={holdout_frac}, Train Rows: {len(X_train)}, Val Rows: {len(X_val)}')
 
-        self._train_multi_and_ensemble(X_train, y_train, X_val, y_val, X_unlabeled=X_unlabeled, hyperparameters=hyperparameters,
-                                       feature_prune=feature_prune,
-                                       num_stack_levels=num_stack_levels, time_limit=time_limit, core_kwargs=core_kwargs)
+            self._train_multi_and_ensemble(X_train, y_train, X_val, y_val, X_unlabeled=X_unlabeled, hyperparameters=hyperparameters,
+                                           feature_prune=feature_prune,
+                                           num_stack_levels=num_stack_levels, time_limit=time_limit, core_kwargs=core_kwargs)
 
     def get_models_distillation(self, hyperparameters, **kwargs):
         path = kwargs.pop('path', self.path)