Nigel
/
mindspore
forked from mindspore-Ecosystem/mindspore
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

modify addcdiv and addcmul

This commit is contained in:
changzherui 2023-03-04 23:21:06 +08:00
parent 934e0b6577
commit b0013fe301
9 changed files with 82 additions and 94 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
docs/api/api_python/mindspore/Tensor/mindspore.Tensor.addcdiv.rst
View File

@ -1,6 +1,6 @@
mindspore.Tensor.addcdiv
========================
.. py:method:: mindspore.Tensor.addcdiv(x1, x2, value)
.. py:method:: mindspore.Tensor.addcdiv(tensor1, tensor2, value=1)
详情请参考 :func:`mindspore.ops.addcdiv`。

2
docs/api/api_python/mindspore/Tensor/mindspore.Tensor.addcmul.rst
View File

@ -1,6 +1,6 @@
mindspore.Tensor.addcmul
========================
.. py:method:: mindspore.Tensor.addcmul(x1, x2, value)
.. py:method:: mindspore.Tensor.addcmul(tensor1, tensor2, value=1)
详情请参考 :func:`mindspore.ops.addcmul`。

24
docs/api/api_python/ops/mindspore.ops.func_addcdiv.rst
View File

@ -1,24 +1,24 @@
mindspore.ops.addcdiv
======================
.. py:function:: mindspore.ops.addcdiv(input_data, x1, x2, value)
.. py:function:: mindspore.ops.addcdiv(input, tensor1, tensor2, value=1)
执行Tensor `x1` 与Tensor `x2` 的逐元素除法，将结果乘以标量值 `value` ，并将其添加到 `input_data` 中。
执行Tensor `tensor1` 与Tensor `tensor2` 的逐元素除法，将结果乘以标量值 `value` ，并将其添加到 `input` 中。
.. math::
y[i] = input\_data[i] + value[i] * (x1[i] / x2[i])
y[i] = input[i] + value[i] * (tensor1[i] / tensor2[i])
参数：
- **input_data** (Tensor) - 要添加的Tensor。
- **x1** (Tensor) - 分子Tensor。
- **x2** (Tensor) - 分母Tensor。
- **value** (Tensor) - Tensor x1/x2的乘数。
- **input** (Tensor) - 要添加的Tensor。
- **tensor1** (Tensor) - 分子Tensor。
- **tensor2** (Tensor) - 分母Tensor。
- **value** (Union[Tensor, Number]) - tensor1 / tensor2的乘数。默认值：1。
返回：
Tensor，具有与x1/x2相同的shape和dtype。
Tensor，具有与tensor1/tensor2相同的shape和dtype。
异常：
- **TypeError** - 如果 `x1` 、 `x2` 、 `value` 、 `input_data` 不是Tensor。
- **ValueError** - 如果无法将 `x1` 广播到 `x2` 。
- **ValueError** - 如果无法将 `value` 广播到 `x1/x2` 。
- **ValueError** - 如果无法将 `input_data` 广播到 `value*(x1/x2)` 。
- **TypeError** - 如果 `tensor1` 、 `tensor2`、 `input` 不是Tensor。
- **ValueError** - 如果无法将 `tensor1` 广播到 `tensor2` 。
- **ValueError** - 如果无法将 `value` 广播到 `tensor1/tensor2` 。
- **ValueError** - 如果无法将 `input` 广播到 `value*(tensor1/tensor2)` 。

28
docs/api/api_python/ops/mindspore.ops.func_addcmul.rst
View File

@ -1,27 +1,27 @@
mindspore.ops.addcmul
======================
.. py:function:: mindspore.ops.addcmul(input_data, x1, x2, value)
.. py:function:: mindspore.ops.addcmul(input, tensor1, tensor2, value=1)
执行Tensor `x1` 与Tensor `x2` 的逐元素乘积，将结果乘以标量值 `value` ，并将其添加到 `input_data` 中。
执行Tensor `tensor1` 与Tensor `tensor2` 的逐元素乘积，将结果乘以标量值 `value` ，并将其添加到 `input` 中。
.. math::
output[i] = input\_data[i] + value[i] * (x1[i] * x2[i])
output[i] = input[i] + value[i] * (tensor1[i] * tensor2[i])
参数：
- **input_data** (Tensor) - 要添加的Tensor。
- **x1** (Tensor) - 要乘以的Tensor。
- **x2** (Tensor) - 要乘以的Tensor。
- **value** (Tensor) - Tensor x1*x2的乘数。
- **input** (Tensor) - 要添加的Tensor。
- **tensor1** (Tensor) - 要乘以的Tensor。
- **tensor2** (Tensor) - 要乘以的Tensor。
- **value** (Union[Tensor, Number]) - tensor1 * tensor2的乘数。默认值：1。
返回：
Tensor，具有与x1*x2相同的shape和dtype。
Tensor，具有与tensor1*tensor2相同的shape和dtype。
异常：
- **TypeError** - 如果 `x1` 、 `x2` 、 `value` 、 `input_data` 不是Tensor。
- **TypeError** - 如果 `input_data` 的dtype不是：float32、float16、int32之一。
- **TypeError** - 如果 `x1` 或 `x2` 的dtype不是：float32、float16、int32之一.
- **TypeError** - 如果 `tensor1` 、 `tensor2`、 `input` 不是Tensor。
- **TypeError** - 如果 `input` 的dtype不是：float32、float16、int32之一。
- **TypeError** - 如果 `tensor1` 或 `tensor2` 的dtype不是：float32、float16、int32之一.
- **TypeError** - 如果 `value` 的dtype不是：float32、float16、int32之一。
- **ValueError** - 如果无法将 `x1` 广播到 `x2` 。
- **ValueError** - 如果无法将 `value` 广播到 `x1` * `x2` 。
- **ValueError** - 如果无法将 `input_data` 广播到 `value*(x1*x2)` 。
- **ValueError** - 如果无法将 `tensor1` 广播到 `tensor2` 。
- **ValueError** - 如果无法将 `value` 广播到 `tensor1` * `tensor2` 。
- **ValueError** - 如果无法将 `input` 广播到 `value*(tensor1*tensor2)` 。

34
mindspore/python/mindspore/_extends/parse/standard_method.py
View File

@ -57,8 +57,6 @@ _reduce_sum_default = P.ReduceSum()
_reduce_sum_keepdims = P.ReduceSum(True)
_mean_keepdims = P.ReduceMean(True)
_csr_mm = _csr_ops.CSRMM()
_addcdiv = P.Addcdiv()
_addcmul = P.Addcmul()
itemsize_map = {mstype.bool_: 1, mstype.int8: 1, mstype.uint8: 1,
mstype.float16: 2, mstype.int16: 2, mstype.uint16: 2,
@ -129,38 +127,38 @@ def prod(x, axis=(), keep_dims=False):
return F.prod(x, axis, keep_dims)
def addcdiv(input_data, x1, x2, value):
def addcdiv(input, tensor1, tensor2, value=1):
"""
Performs the element-wise division of tensor x1 by tensor x2,
Performs the element-wise division of tensor tensor1 by tensor tensor2,
multiply the result by the scalar value and add it to input_data.
Args:
input_data (Tensor): The tensor to be added.
x1 (Tensor): The numerator tensor.
x2 (Tensor): The denominator tensor.
value (Tensor): The multiplier for tensor x1/x2.
input (Tensor): The tensor to be added.
tensor1 (Tensor): The numerator tensor.
tensor1 (Tensor): The denominator tensor.
value (Union[Tensor, Number]): The multiplier for tensor1/tensor2. Default: 1.
Returns:
Tensor, has the same shape and dtype as x1/x2.
Tensor, has the same shape and dtype as tensor1 / tensor2.
"""
return _addcdiv(input_data, x1, x2, value)
return F.addcdiv(input, tensor1, tensor2, value)
def addcmul(input_data, x1, x2, value):
def addcmul(input, tensor1, tensor2, value=1):
"""
Performs the element-wise product of tensor x1 and tensor x2,
Performs the element-wise product of tensor tensor1 and tensor tensor2,
multiply the result by the scalar value and add it to input_data.
Args:
input_data (Tensor): The tensor to be added.
x1 (Tensor): The tensor to be multiplied.
x2 (Tensor): The tensor to be multiplied.
value (Tensor): The multiplier for tensor x1*x2.
input (Tensor): The tensor to be added.
tensor1 (Tensor): The tensor to be multiplied.
tensor2 (Tensor): The tensor to be multiplied.
value (Union[Tensor, Number]): The multiplier for tensor1*tensor2. Default: 1.
Returns:
Tensor, has the same shape and dtype as x1*x2.
Tensor, has the same shape and dtype as tensor1 * tensor2.
"""
return _addcmul(input_data, x1, x2, value)
return F.addcmul(input, tensor1, tensor2, value)
def all_(x, axis=(), keep_dims=False):

9
mindspore/python/mindspore/common/tensor.py
View File

@ -795,20 +795,19 @@ class Tensor(Tensor_, metaclass=_TensorMeta):
self._init_check()
Tensor_._flush_from_cache(self)
def addcdiv(self, x1, x2, value):
def addcdiv(self, tensor1, tensor2, value=1):
r"""
For details, please refer to :func:`mindspore.ops.addcdiv`.
"""
self._init_check()
return tensor_operator_registry.get('addcdiv')()(self, x1, x2, value)
return tensor_operator_registry.get('addcdiv')()(self, tensor1, tensor2, value)
def addcmul(self, x1, x2, value):
def addcmul(self, tensor1, tensor2, value=1):
r"""
For details, please refer to :func:`mindspore.ops.addcmul`.
"""
self._init_check()
return tensor_operator_registry.get('addcmul')()(self, x1, x2, value)
return tensor_operator_registry.get('addcmul')()(self, tensor1, tensor2, value)
def add(self, y):
r"""

54
mindspore/python/mindspore/ops/function/math_func.py
View File

@ -307,28 +307,28 @@ def add(x, y):
return tensor_add(x, y)
def addcdiv(input_data, x1, x2, value):
def addcdiv(input, tensor1, tensor2, value=1):
r"""
Performs the element-wise division of tensor x1 by tensor x2,
Performs the element-wise division of tensor tensor1 by tensor tensor2,
multiply the result by the scalar value and add it to input_data.
.. math::
y[i] = input\_data[i] + value[i] * (x1[i] / x2[i])
y[i] = input[i] + value[i] * (tensor1[i] / tensor2[i])
Args:
input_data (Tensor): The tensor to be added.
x1 (Tensor): The numerator tensor.
x2 (Tensor): The denominator tensor.
value (Tensor): The multiplier for tensor x1/x2.
input (Tensor): The tensor to be added.
tensor1 (Tensor): The numerator tensor.
tensor2 (Tensor): The denominator tensor.
value (Union[Tensor, Number]): The multiplier for tensor1/tensor2. Default: 1.
Returns:
Tensor, has the same shape and dtype as x1/x2.
Tensor, has the same shape and dtype as tensor1/tensor2.
Raises:
TypeError: If dtype of `x1`, `x2`, `value`, `input_data` is not tensor.
ValueError: If `x1` could not be broadcast to a tensor with shape of `x2`.
ValueError: If `value` could not be broadcast to tensors with shapes of `x1/x2`.
ValueError: If `input_data` could not be broadcast to tensors with shapes of `value*(x1/x2)`.
TypeError: If dtype of `tensor1`, `tensor2`, `input` is not tensor.
ValueError: If `tensor1` could not be broadcast to a tensor with shape of `tensor2`.
ValueError: If `value` could not be broadcast to tensors with shapes of `tensor1/tensor2`.
ValueError: If `input` could not be broadcast to tensors with shapes of `value*(tensor1/tensor2)`.
Supported Platforms:
``Ascend`` ``GPU`` ``CPU``
@ -342,34 +342,34 @@ def addcdiv(input_data, x1, x2, value):
>>> print(y)
[1.25 1.6666667 2.5 5. ]
"""
return _get_cache_prim(P.Addcdiv)()(input_data, x1, x2, value)
return _get_cache_prim(P.Addcdiv)()(input, tensor1, tensor2, Tensor(value))
def addcmul(input_data, x1, x2, value):
def addcmul(input, tensor1, tensor2, value=1):
r"""
Performs the element-wise product of tensor x1 and tensor x2,
Performs the element-wise product of tensor tensor1 and tensor tensor2,
multiply the result by the scalar value and add it to input_data.
.. math::
output[i] = input\_data[i] + value[i] * (x1[i] * x2[i])
output[i] = input[i] + value[i] * (tensor1[i] * tensor2[i])
Args:
input_data (Tensor): The tensor to be added.
x1 (Tensor): The tensor to be multiplied.
x2 (Tensor): The tensor to be multiplied.
value (Tensor): The multiplier for tensor x1*x2.
input (Tensor): The tensor to be added.
tensor1 (Tensor): The tensor to be multiplied.
tensor2 (Tensor): The tensor to be multiplied.
value (Union[Tensor, Number]): The multiplier for tensor1*tensor2. Default: 1.
Returns:
Tensor, has the same shape and dtype as x1*x2.
Raises:
TypeError: If dtype of `x1`, `x2`, `value`, `input_data` is not tensor.
TypeError: If dtype of `input_data` is not one of: float32, float16, int32.
TypeError: If dtype of `x1` or `x2` is not one of: float32, float16, int32.
TypeError: If dtype of `tensor1`, `tensor2`, `input` is not Tensor.
TypeError: If dtype of `input` is not one of: float32, float16, int32.
TypeError: If dtype of `tensor1` or `tensor2` is not one of: float32, float16, int32.
TypeError: If dtype of `value` is not one of: float32, float16, int32.
ValueError: If `x1` could not be broadcast to a tensor with shape of `x2`.
ValueError: If `value` could not be broadcast to tensors with shapes of `x1` * `x2`.
ValueError: If `input_data` could not be broadcast to tensors with shapes of `value*(x1*x2)`.
ValueError: If `tensor1` could not be broadcast to a tensor with shape of `tensor2`.
ValueError: If `value` could not be broadcast to tensors with shapes of `tensor1` * `tensor2`.
ValueError: If `input` could not be broadcast to tensors with shapes of `value*(tensor1*tensor2)`.
Supported Platforms:
``Ascend`` ``GPU`` ``CPU``
@ -385,7 +385,7 @@ def addcmul(input_data, x1, x2, value):
[ 3. 5. 7.]
[ 4. 7. 10.]]
"""
return _get_cache_prim(P.Addcmul)()(input_data, x1, x2, value)
return _get_cache_prim(P.Addcmul)()(input, tensor1, tensor2, Tensor(value))
def angle(x):

12
tests/st/ops/gpu/test_addcdiv_op.py
View File

@ -76,14 +76,10 @@ def test_addcdiv_float64_pynative_value(type_s=np.float64):
).astype(type_s)
)
x2 = Tensor(np.array([[0.73312728], [0.50139652]]).astype(type_s))
value = Tensor(
np.array(
[
[0.6975477, 0.89641169, -0.16985319, -0.6640372],
[0.79931823, -1.65808474, 0.17895249, -1.41405968],
]
).astype(type_s)
)
value = np.array([
[0.6975477, 0.89641169, -0.16985319, -0.6640372],
[0.79931823, -1.65808474, 0.17895249, -1.41405968],
])
addcdiv = P.Addcdiv()
output = addcdiv(input_data, x1, x2, value)
output_ms = output.asnumpy()

11
tests/st/ops/gpu/test_addcmul_op.py
View File

@ -77,14 +77,9 @@ def test_addcmul_float64_pynative_value(type_s=np.float64):
).astype(type_s)
)
x2 = Tensor(np.array([[0.67554175], [0.25244115]]).astype(type_s))
value = Tensor(
np.array(
[
[-0.99526738, -2.57715965, -0.73273605, 0.02546449],
[1.34531189, -0.97279591, 2.4665573, -1.18791833],
]
).astype(type_s)
)
value = np.array([
[-0.99526738, -2.57715965, -0.73273605, 0.02546449],
[1.34531189, -0.97279591, 2.4665573, -1.18791833],])
addcmul = P.Addcmul()
output = addcmul(input_data, x1, x2, value)
output_ms = output.asnumpy()