【SCU】【Paddle Tensor No.25】新增paddle.vecdot , paddle.linalg.vecdot

python/paddle/__init__.py

@@ -241,6 +241,7 @@
     t_,
     transpose,
     transpose_,
+    vecdot,
 )
 from .tensor.logic import (
     allclose,
@@ -1213,5 +1214,6 @@
     'to_dlpack',
     'inf',
     'newaxis',
+    'vecdot',
     'nan',
 ]

python/paddle/linalg.py

@@ -46,6 +46,7 @@
     svd,
     svd_lowrank,
     triangular_solve,
+    vecdot,
     vector_norm,
 )
 
@@ -54,6 +55,7 @@
     'cholesky_inverse',
     'norm',
     'matrix_norm',
+    'vecdot',
     'vector_norm',
     'cond',
     'cov',

python/paddle/tensor/__init__.py

@@ -102,6 +102,7 @@
     t_,
     transpose,
     transpose_,
+    vecdot,
 )
 from .logic import (  # noqa: F401
     allclose,

python/paddle/tensor/linalg.py

@@ -1867,6 +1867,49 @@ def dot(x: Tensor, y: Tensor, name: str | None = None) -> Tensor:
         return out
 
 
+def vecdot(
+    x: Tensor,
+    y: Tensor,
+    axis: int = -1,
+    name: str | None = None,
+) -> Tensor:
+    """
+    Computes the dot product of two tensors along a specified axis.
+
+    This function multiplies two tensors element-wise and sums them along a specified axis to compute their dot product. It supports tensors of any dimensionality, including 0-D tensors, as long as the shapes of `x` and `y` are broadcastable along the specified axis.
+
+    Args:
+        x (Tensor): The first input tensor. It should be a tensor with dtype of float32, float64, int32, int64, complex64, or complex128.
+        y (Tensor): The second input tensor. Its shape must be broadcastable with `x` along the specified `axis`, and it must have the same dtype as `x`.
+        axis (int, optional): The axis along which to compute the dot product. Default is -1, which indicates the last axis.
+        name (str|None, optional): Name of the output. Default is None. It's used to print debug info for developers. Details: :ref:`api_guide_Name`
+
+    Returns:
+        Tensor: A tensor containing the dot product of `x` and `y` along the specified axis.
+
+    Examples:
+
+        .. code-block:: python
+
+            >>> import paddle
+            >>> x = paddle.to_tensor([1, 2, 3], dtype='float32')
+            >>> y = paddle.to_tensor([4, 5, 6], dtype='float32')
+            >>> result = paddle.linalg.vecdot(x, y)
+            >>> print(result)
+            Tensor(shape=[], dtype=float32, place=Place(cpu), stop_gradient=True,
+               [32.0])
+
+            >>> x2 = paddle.to_tensor([[1, 2, 3], [4, 5, 6]], dtype='float32')
+            >>> y2 = paddle.to_tensor([[1, 2, 3], [4, 5, 6]], dtype='float32')
+            >>> result2 = paddle.linalg.vecdot(x2, y2, axis=1)
+            >>> print(result2)
+            Tensor(shape=[2], dtype=float32, place=Place(cpu), stop_gradient=True,
+               [14.0, 77.0])
+    """
+    out = (x.conj() * y).sum(axis=axis)
+    return out
+
+
 def cov(
     x: Tensor,
     rowvar: bool = True,

test/legacy_test/test_linalg_vecdot.py

@@ -0,0 +1,274 @@
+#   Copyright (c) 2024 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 sys
+import unittest
+
+import numpy as np
+
+import paddle
+from paddle.base import core
+
+if sys.platform == 'win32':
+    RTOL = {'float32': 1e-02, 'float64': 1e-04}
+    ATOL = {'float32': 1e-02, 'float64': 1e-04}
+else:
+    RTOL = {'float32': 1e-06, 'float64': 1e-15}
+    ATOL = {'float32': 1e-06, 'float64': 1e-15}
+
+
+class VecDotTestCase(unittest.TestCase):
+    def setUp(self):
+        self.init_config()
+        self.generate_input()
+        self.generate_expected_output()
+        self.places = [paddle.CPUPlace()]
+        if paddle.is_compiled_with_cuda():
+            self.places.append(paddle.CUDAPlace(0))
+
+    def generate_input(self):
+        np.random.seed(123)
+        self.x = np.random.random(self.input_shape).astype(self.dtype)
+        self.y = np.random.random(self.input_shape).astype(self.dtype)
+
+    def generate_expected_output(self):
+        self.expected_output = np.sum(self.x * self.y, axis=self.axis)
+
+    def init_config(self):
+        self.dtype = 'float64'
+        self.input_shape = (3, 4)
+        self.axis = -1
+
+    def test_dygraph(self):
+        for place in self.places:
+            paddle.disable_static(place)
+            x_tensor = paddle.to_tensor(self.x, dtype=self.dtype, place=place)
+            y_tensor = paddle.to_tensor(self.y, dtype=self.dtype, place=place)
+            result = paddle.vecdot(x_tensor, y_tensor, axis=self.axis)
+
+            np.testing.assert_allclose(
+                result.numpy(),
+                self.expected_output,
+                rtol=RTOL[self.dtype],
+                atol=ATOL[self.dtype],
+            )
+
+    def test_static(self):
+        paddle.enable_static()
+        for place in self.places:
+            with paddle.static.program_guard(
+                paddle.static.Program(), paddle.static.Program()
+            ):
+                x = paddle.static.data(
+                    name="x", shape=self.input_shape, dtype=self.dtype
+                )
+                y = paddle.static.data(
+                    name="y", shape=self.input_shape, dtype=self.dtype
+                )
+
+                result = paddle.vecdot(x, y, axis=self.axis)
+                exe = paddle.static.Executor(place)
+                output = exe.run(
+                    feed={"x": self.x, "y": self.y},
+                    fetch_list=[result],
+                )[0]
+
+            np.testing.assert_allclose(
+                output,
+                self.expected_output,
+                rtol=RTOL[self.dtype],
+                atol=ATOL[self.dtype],
+            )
+
+
+class VecDotTestCaseFloat32(VecDotTestCase):
+    def init_config(self):
+        self.dtype = 'float32'
+        self.input_shape = (3, 4)
+        self.axis = -1
+
+
+class VecDotTestCaseHigherDim(VecDotTestCase):
+    def init_config(self):
+        self.dtype = 'float64'
+        self.input_shape = (2, 3, 4)
+        self.axis = -1
+
+
+class VecDotTestCaseAxis(VecDotTestCase):
+    def init_config(self):
+        self.dtype = 'float64'
+        self.input_shape = (3, 4, 5)
+        self.axis = 1
+
+
+class VecDotTestCaseError(unittest.TestCase):
+    def test_axis_mismatch(self):
+        with self.assertRaises(ValueError):
+            x = paddle.rand([3, 4], dtype="float32")
+            y = paddle.rand([3, 5], dtype="float32")
+            paddle.vecdot(x, y, axis=-1)
+
+    @unittest.skipIf(
+        core.is_compiled_with_xpu(),
+        "Skip XPU for not support uniform(dtype=int)",
+    )
+    def test_dtype_mismatch(self):
+        with self.assertRaises(TypeError):
+            x = paddle.rand([3, 4], dtype="float32")
+            y = paddle.rand([3, 4], dtype="int32")
+            paddle.vecdot(x, y, axis=-1)
+
+
+@unittest.skipIf(
+    core.is_compiled_with_xpu(),
+    "Skip XPU for not support uniform(dtype=int)",
+)
+class VecDotTestCaseComplex(unittest.TestCase):
+    def run_test_dynamic(self):
+        paddle.disable_static()
+        x = paddle.to_tensor(
+            [[1 + 2j, 3 + 4j], [5 + 6j, 7 + 8j]], dtype="complex64"
+        )
+        y = paddle.to_tensor(
+            [[9 + 1j, 8 + 2j], [7 + 3j, 6 + 4j]], dtype="complex64"
+        )
+        result = paddle.vecdot(x, y, axis=-1)
+        expected = np.sum((x.numpy().conj() * y.numpy()), axis=-1)
+        np.testing.assert_allclose(
+            result.numpy(), expected, rtol=1e-5, atol=1e-5
+        )
+
+    def run_test_static(self):
+        paddle.enable_static()
+        place = paddle.CPUPlace()
+        with paddle.static.program_guard(paddle.static.Program()):
+            x = paddle.static.data(name="x", shape=[2, 2], dtype="complex64")
+            y = paddle.static.data(name="y", shape=[2, 2], dtype="complex64")
+            result = paddle.vecdot(x, y, axis=-1)
+            exe = paddle.static.Executor(place)
+            output = exe.run(
+                feed={
+                    "x": np.array([[1 + 2j, 3 + 4j], [5 + 6j, 7 + 8j]]).astype(
+                        "complex64"
+                    ),
+                    "y": np.array([[9 + 1j, 8 + 2j], [7 + 3j, 6 + 4j]]).astype(
+                        "complex64"
+                    ),
+                },
+                fetch_list=[result],
+            )[0]
+        expected = np.sum(
+            np.conj(np.array([[1 + 2j, 3 + 4j], [5 + 6j, 7 + 8j]])).astype(
+                "complex64"
+            )
+            * np.array([[9 + 1j, 8 + 2j], [7 + 3j, 6 + 4j]]).astype(
+                "complex64"
+            ),
+            axis=-1,
+        )
+        np.testing.assert_allclose(output, expected, rtol=1e-5, atol=1e-5)
+
+    def test_complex_conjugate(self):
+        self.run_test_dynamic()
+        self.run_test_static()
+
+
+@unittest.skipIf(
+    core.is_compiled_with_xpu(),
+    "Skip XPU for not support uniform(dtype=int)",
+)
+class VecDotTestCaseTypePromotion1(unittest.TestCase):
+    def test_float32_float64_promotion(self):
+        paddle.disable_static()
+        x = paddle.to_tensor([[1.0, 2.0], [3.0, 4.0]], dtype="float32")
+        y = paddle.to_tensor([[5.0, 6.0], [7.0, 8.0]], dtype="float64")
+        result = paddle.vecdot(x, y, axis=-1)
+
+        expected = np.sum(x.numpy().astype("float64") * y.numpy(), axis=-1)
+        np.testing.assert_allclose(
+            result.numpy(), expected, rtol=1e-6, atol=1e-6
+        )
+
+
+class VecDotTestCaseTypePromotion2(unittest.TestCase):
+    def test_float64_complex64_promotion(self):
+        paddle.disable_static()
+        x = paddle.to_tensor([[1.0, 2.0], [3.0, 4.0]], dtype="float64")
+        y = paddle.to_tensor(
+            [[5 + 6j, 7 + 8j], [9 + 1j, 2 + 3j]], dtype="complex64"
+        )
+        result = paddle.vecdot(x, y, axis=-1)
+
+        expected = np.sum(x.numpy().astype("complex64") * y.numpy(), axis=-1)
+        np.testing.assert_allclose(
+            result.numpy(), expected, rtol=1e-5, atol=1e-5
+        )
+
+
+class VecDotTestCaseBroadcast0DTensor(unittest.TestCase):
+    def test_0d_tensor_broadcast(self):
+        paddle.disable_static()
+        x = paddle.to_tensor(2.0, dtype="float32")
+        y = paddle.to_tensor(3.0, dtype="float32")
+        result = paddle.vecdot(x, y)
+
+        expected = x.numpy() * y.numpy()
+        np.testing.assert_allclose(
+            result.numpy(), expected, rtol=1e-6, atol=1e-6
+        )
+
+
+class VecDotTestCaseBroadcast1DTensor(unittest.TestCase):
+    def test_1d_tensor_broadcast(self):
+        paddle.disable_static()
+        x = paddle.to_tensor([1.0, 2.0, 3.0], dtype="float32")
+        y = paddle.to_tensor([4.0, 5.0, 6.0], dtype="float32")
+        result = paddle.vecdot(x, y)
+
+        expected = np.dot(x.numpy(), y.numpy())
+        np.testing.assert_allclose(
+            result.numpy(), expected, rtol=1e-6, atol=1e-6
+        )
+
+
+class VecDotTestCaseBroadcast1DNDTensor(unittest.TestCase):
+    def test_1d_nd_tensor_broadcast(self):
+        paddle.disable_static()
+        x = paddle.to_tensor([1.0, 2.0], dtype="float32")
+        y = paddle.to_tensor([[3.0, 4.0], [5.0, 6.0]], dtype="float32")
+        result = paddle.vecdot(x, y, axis=-1)
+
+        expected = np.sum(x.numpy() * y.numpy(), axis=-1)
+        np.testing.assert_allclose(
+            result.numpy(), expected, rtol=1e-6, atol=1e-6
+        )
+
+
+class VecDotTestCaseBroadcastNDTensor(unittest.TestCase):
+    def test_nd_nd_tensor_broadcast(self):
+        paddle.disable_static()
+        x = paddle.to_tensor([[1.0, 2.0], [3.0, 4.0]], dtype="float32")
+        y = paddle.to_tensor([5.0, 6.0], dtype="float32")
+        result = paddle.vecdot(x, y, axis=-1)
+
+        expected = np.sum(x.numpy() * y.numpy(), axis=-1)
+        np.testing.assert_allclose(
+            result.numpy(), expected, rtol=1e-6, atol=1e-6
+        )
+
+
+if __name__ == '__main__':
+    unittest.main()