Fix compatibility with NumPy v1.24 (#3323)

* Fix

* More fixes

* Use latest scikit-learn

.github/workflows/platform-ci.yml

@@ -72,7 +72,7 @@ jobs:
 
           source ./ci/rewrite-cov-config.sh
 
-          pip install numpy scipy cython "scikit-learn<1.2.0"
+          pip install numpy scipy cython
 
           pip install -e ".[dev,extra]"
 
@@ -104,11 +104,6 @@ jobs:
             fi
             if [ -n "$WITH_RAY" ] || [ -n "$WITH_RAY_DAG" ] || [ -n "$WITH_RAY_DEPLOY" ]; then
               pip install "xgboost_ray" "protobuf<4"
-              # Future PR will Unpin numpy.
-              pip install "numpy<1.24"
-              # Use standard ray releases when ownership bug is fixed
-              # pip uninstall -y ray
-              # pip install https://s3-us-west-2.amazonaws.com/ray-wheels/master/c03d0432f3bb40f3c597b7fc450870ba5e34ad56/ray-3.0.0.dev0-cp38-cp38-manylinux2014_x86_64.whl
               # Ray Datasets need pyarrow>=6.0.1
               pip install "pyarrow>=6.0.1"
             fi

azure-pipelines.yml

@@ -60,7 +60,7 @@ jobs:
         source ./ci/rewrite-cov-config.sh
 
         if [[ "$(mars.test.module)" == "learn" ]]; then
-          pip install numpy\<1.20.0 scipy cython scikit-learn\<1.2.0
+          pip install numpy\<1.20.0 scipy cython
         else
           pip install numpy scipy cython
         fi

mars/learn/datasets/tests/test_samples_generator.py

@@ -110,7 +110,7 @@ def test_make_classification_informative_features(setup):
         (2, [1 / 2] * 2, 2),
         (2, [3 / 4, 1 / 4], 2),
         (10, [1 / 3] * 3, 10),
-        (np.int(64), [1], 1),
+        (np.int_(64), [1], 1),
     ]:
         n_classes = len(weights)
         n_clusters = n_classes * n_clusters_per_class

mars/learn/utils/tests/test_multiclass.py

@@ -117,7 +117,7 @@
         [()],
         # sequence of sequences that weren't supported even before deprecation
         np.array([np.array([]), np.array([1, 2, 3])], dtype=object),
-        [np.array([]), np.array([1, 2, 3])],
+        # [np.array([]), np.array([1, 2, 3])], # deprecated in numpy v1.24
         [{1, 2, 3}, {1, 2}],
         [frozenset([1, 2, 3]), frozenset([1, 2])],
         # and also confusable as sequences of sequences
@@ -222,7 +222,7 @@ def test_is_multilabel(setup):
     raws = [
         [[1, 2]],
         [0, 1, 0, 1],
-        [[1], [0, 2], []],
+        # [[1], [0, 2], []], # deprecated in numpy v1.24
         np.array([[1, 0], [0, 0]]),
         np.array([[1], [0], [0]]),
         np.array([[1, 0, 0]]),

mars/tensor/__init__.py

@@ -334,14 +334,14 @@
     character,
     generic,
     flexible,
-    int_,
-    bool_,
-    float_,
+    int_ as int,
+    bool_ as bool,
+    float_ as float,
     cfloat,
     bytes_,
     unicode_,
     void,
-    object_,
+    object_ as object,
     intc,
     intp,
     int8,
@@ -371,20 +371,6 @@
 from .fetch import TensorFetch, TensorFetchShuffle
 from . import ufunc
 
-try:
-    import warnings
-
-    # suppress numpy warnings on types
-    with warnings.catch_warnings():
-        warnings.simplefilter("ignore", DeprecationWarning)
-        warnings.simplefilter("ignore", FutureWarning)
-        # noinspection PyUnresolvedReferences
-        from numpy import object, int, bool, float
-except ImportError:  # pragma: no cover
-    pass
-finally:
-    del warnings
-
 del (
     TensorFuseChunk,
     TensorCpFuseChunk,

mars/tests/test_session.py

@@ -188,7 +188,7 @@ def test_array_protocol(setup):
 
     arr2 = mt.ones((10, 20))
 
-    result = np.asarray(arr2, mt.bool_)
+    result = np.asarray(arr2, mt.bool)
     np.testing.assert_array_equal(result, np.ones((10, 20), dtype=np.bool_))
 
     arr3 = mt.ones((10, 20)).sum()