【Hackathon 5th No.14】Add combinations API to Paddle

python/paddle/__init__.py

@@ -420,6 +420,7 @@
     polygamma_,
     hypot,
     hypot_,
+    combinations,
 )
 
 from .tensor.random import (
@@ -933,4 +934,5 @@
     'index_fill',
     "index_fill_",
     'diagonal_scatter',
+    'combinations',
 ]

python/paddle/tensor/__init__.py

@@ -228,6 +228,7 @@
     ceil_,
     clip,
     clip_,
+    combinations,
     conj,
     cos,
     cos_,
@@ -762,6 +763,7 @@
     'atleast_2d',
     'atleast_3d',
     'diagonal_scatter',
+    "combinations",
 ]
 
 # this list used in math_op_patch.py for magic_method bind

python/paddle/tensor/math.py

@@ -7071,3 +7071,69 @@ def hypot_(x, y, name=None):
 
     out = x.pow_(2).add_(y.pow(2)).sqrt_()
     return out
+
+
+def combinations(x, r=2, with_replacement=False, name=None):
+    """
+
+    Compute combinations of length r of the given tensor. The behavior is similar to python's itertools.combinations
+    when with_replacement is set to False, and itertools.combinations_with_replacement when with_replacement is set to True.
+
+    Args:
+        x (Tensor): 1-D input Tensor, the data type is float16, float32, float64, int32 or int64.
+        r (int, optional):  number of elements to combine, default value is 2.
+        with_replacement (bool, optional):  whether to allow duplication in combination, default value is False.
+        name (str, optional): Name for the operation (optional, default is None).For more information, please refer to :ref:`api_guide_Name`.
+
+    Returns:
+        out (Tensor). Tensor concatenated by combinations, same dtype with x.
+
+    Examples:
+        .. code-block:: python
+
+            >>> import paddle
+            >>> x = paddle.to_tensor([1, 2, 3], dtype='int32')
+            >>> res = paddle.combinations(x)
+            >>> print(res)
+            Tensor(shape=[3, 2], dtype=int32, place=Place(gpu:0), stop_gradient=True,
+                   [[1, 2],
+                    [1, 3],
+                    [2, 3]])
+
+    """
+    if len(x.shape) != 1:
+        raise TypeError(f"Expect a 1-D vector, but got x shape {x.shape}")
+    if not isinstance(r, int) or r < 0:
+        raise ValueError(f"Expect a non-negative int, but got r={r}")
+
+    if r == 0:
+        return paddle.empty(shape=[0], dtype=x.dtype)
+
+    if (r > x.shape[0] and not with_replacement) or (
+        x.shape[0] == 0 and with_replacement
+    ):
+        return paddle.empty(shape=[0, r], dtype=x.dtype)
+
+    if r > 1:
+        t_l = [x for i in range(r)]
+        grids = paddle.meshgrid(t_l)
+    else:
+        grids = [x]
+    num_elements = x.numel()
+    t_range = paddle.arange(num_elements, dtype='int64')
+    if r > 1:
+        t_l = [t_range for i in range(r)]
+        index_grids = paddle.meshgrid(t_l)
+    else:
+        index_grids = [t_range]
+    mask = paddle.full(x.shape * r, True, dtype='bool')
+    if with_replacement:
+        for i in range(r - 1):
+            mask *= index_grids[i] <= index_grids[i + 1]
+    else:
+        for i in range(r - 1):
+            mask *= index_grids[i] < index_grids[i + 1]
+    for i in range(r):
+        grids[i] = grids[i].masked_select(mask)
+
+    return paddle.stack(grids, 1)

test/legacy_test/test_combinations.py

@@ -0,0 +1,150 @@
+# Copyright (c) 2023 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+import random
+import unittest
+from itertools import combinations, combinations_with_replacement
+
+import numpy as np
+
+import paddle
+from paddle.base import Program
+
+paddle.enable_static()
+
+
+def convert_combinations_to_array(x, r=2, with_replacement=False):
+    if r == 0:
+        return np.array([]).astype(x.dtype)
+    if with_replacement:
+        combs = combinations_with_replacement(x, r)
+    else:
+        combs = combinations(x, r)
+    combs = list(combs)
+    res = []
+    for i in range(len(combs)):
+        res.append(list(combs[i]))
+    if len(res) != 0:
+        return np.array(res).astype(x.dtype)
+    else:
+        return np.empty((0, r))
+
+
+class TestCombinationsAPIBase(unittest.TestCase):
+    def setUp(self):
+        self.init_setting()
+        self.modify_setting()
+        self.x_np = np.random.random(self.x_shape).astype(self.dtype_np)
+
+        self.place = ['cpu']
+        if paddle.is_compiled_with_cuda():
+            self.place.append('gpu')
+
+    def init_setting(self):
+        self.dtype_np = 'float64'
+        self.x_shape = [10]
+        self.r = 5
+        self.with_replacement = False
+
+    def modify_setting(self):
+        pass
+
+    def test_static_graph(self):
+        paddle.enable_static()
+        for place in self.place:
+            with paddle.static.program_guard(Program()):
+                x = paddle.static.data(
+                    name="x", shape=self.x_shape, dtype=self.dtype_np
+                )
+                out = paddle.combinations(x, self.r, self.with_replacement)
+                exe = paddle.static.Executor(place=place)
+                feed_list = {"x": self.x_np}
+                pd_res = exe.run(
+                    paddle.static.default_main_program(),
+                    feed=feed_list,
+                    fetch_list=[out],
+                )[0]
+                ref_res = convert_combinations_to_array(
+                    self.x_np, self.r, self.with_replacement
+                )
+                np.testing.assert_allclose(ref_res, pd_res)
+
+    def test_dygraph(self):
+        paddle.disable_static()
+        for place in self.place:
+            paddle.device.set_device(place)
+            x_pd = paddle.to_tensor(self.x_np)
+            pd_res = paddle.combinations(x_pd, self.r, self.with_replacement)
+            ref_res = convert_combinations_to_array(
+                self.x_np, self.r, self.with_replacement
+            )
+            np.testing.assert_allclose(ref_res, pd_res)
+
+    def test_errors(self):
+        def test_input_not_1D():
+            data_np = np.random.random((10, 10)).astype(np.float32)
+            res = paddle.combinations(data_np, self.r, self.with_replacement)
+
+        self.assertRaises(TypeError, test_input_not_1D)
+
+        def test_r_range():
+            res = paddle.combinations(self.x_np, -1, self.with_replacement)
+
+        self.assertRaises(ValueError, test_r_range)
+
+
+class TestCombinationsAPI1(TestCombinationsAPIBase):
+    def modify_setting(self):
+        self.dtype_np = 'int32'
+        self.x_shape = [10]
+        self.r = 1
+        self.with_replacement = True
+
+
+class TestCombinationsAPI2(TestCombinationsAPIBase):
+    def modify_setting(self):
+        self.dtype_np = 'int64'
+        self.x_shape = [10]
+        self.r = 0
+        self.with_replacement = True
+
+
+class TestCombinationsEmpty(unittest.TestCase):
+    def setUp(self):
+        self.place = ['cpu']
+        if paddle.is_compiled_with_cuda():
+            self.place.append('gpu')
+
+    def test_dygraph(self):
+        paddle.disable_static()
+        for place in self.place:
+            paddle.device.set_device(place)
+            a = paddle.rand([3], dtype='float32')
+            c = paddle.combinations(a, r=4)
+            expected = convert_combinations_to_array(a.numpy(), r=4)
+            np.testing.assert_allclose(c, expected)
+
+            # test empty input
+            a = paddle.empty([random.randint(0, 8)])
+            c1 = paddle.combinations(a, r=2)
+            c2 = paddle.combinations(a, r=2, with_replacement=True)
+            expected1 = convert_combinations_to_array(a.numpy(), r=2)
+            expected2 = convert_combinations_to_array(
+                a.numpy(), r=2, with_replacement=True
+            )
+            np.testing.assert_allclose(c1, expected1)
+            np.testing.assert_allclose(c2, expected2)
+
+
+if __name__ == '__main__':
+    unittest.main()