p85971406
/
mindspore
forked from mindspore-Ecosystem/mindspore
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

!47324 modify format 

Merge pull request !47324 from 俞涵/code_docs_1110
This commit is contained in:
i-robot 2022-12-30 01:43:25 +00:00 committed by Gitee
parent aa08c27fb9 19af119abc
commit d5aa0b0588
No known key found for this signature in database
GPG Key ID: 173E9B9CA92EEF8F
31 changed files with 218 additions and 91 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

56
docs/api/api_python/mindspore.communication.rst
View File

@ -2,7 +2,7 @@ mindspore.communication
========================
集合通信接口。
注意，集合通信接口需要预先设置环境变量。
注意，集合通信接口需要先配置好通信环境变量。
针对Ascend设备，用户需要准备rank表，设置rank_id和device_id，详见 `Ascend指导文档 <https://www.mindspore.cn/tutorials/experts/zh-CN/master/parallel/train_ascend.html#准备环节>`_ 。
@ -25,7 +25,6 @@ mindspore.communication
- HCCL的全称是华为集合通信库（Huawei Collective Communication Library）。
- NCCL的全称是英伟达集合通信库（NVIDIA Collective Communication Library）。
- MCCL的全称是MindSpore集合通信库（MindSpore Collective Communication Library）。
- 在运行以下示例之前，用户需要预设通信环境变量，请查看mindspore.communication的文档注释。
参数：
- **backend_name** (str) - 分布式后端的名称，可选HCCL或NCCL。在Ascend硬件平台下，应使用HCCL，在GPU硬件平台下，应使用NCCL。如果未设置则根据硬件平台类型（device_target）自动进行推断，默认值为None。
@ -34,24 +33,32 @@ mindspore.communication
- **TypeError** - 参数 `backend_name` 不是字符串。
- **RuntimeError** - 1）硬件设备类型无效；2）后台服务无效；3）分布式计算初始化失败；4）后端是HCCL的情况下，未设置环境变量 `RANK_ID` 或 `MINDSPORE_HCCL_CONFIG_PATH` 的情况下初始化HCCL服务。
样例：
.. note::
.. include:: ops/mindspore.ops.comm_note.rst
.. py:function:: mindspore.communication.release()
释放分布式资源，例如 `HCCL` 或 `NCCL` 服务。
.. note::
- `release` 方法应该在 `init` 方法之后使用。
- 在运行以下示例之前，用户需要预设通信环境变量，请查看mindspore.communication的文档注释。
异常：
- **RuntimeError** - 在释放分布式资源失败时抛出。
样例：
.. note::
.. include:: ops/mindspore.ops.comm_note.rst
.. py:function:: mindspore.communication.get_rank(group=GlobalComm.WORLD_COMM_GROUP)
在指定通信组中获取当前的设备序号。
.. note::
- `get_rank` 方法应该在 `init` 方法之后使用。
- 在运行以下示例之前，用户需要预设通信环境变量，请查看mindspore.communication的文档注释。
参数：
- **group** (str) - 通信组名称，通常由 `create_group` 方法创建，否则将使用默认组。默认值： `GlobalComm.WORLD_COMM_GROUP` 。
@ -64,13 +71,17 @@ mindspore.communication
- **ValueError** - 在后台不可用时抛出。
- **RuntimeError** - 在 `HCCL` 或 `NCCL` 服务不可用时抛出。
样例：
.. note::
.. include:: ops/mindspore.ops.comm_note.rst
.. py:function:: mindspore.communication.get_group_size(group=GlobalComm.WORLD_COMM_GROUP)
获取指定通信组实例的rank_size。
.. note::
- `get_group_size` 方法应该在 `init` 方法之后使用。
- 在运行以下示例之前，用户需要预设通信环境变量，请查看mindspore.communication的文档注释。
参数：
- **group** (str) - 指定工作组实例（由 create_group 方法创建）的名称，支持数据类型为str，默认值为 `WORLD_COMM_GROUP` 。
@ -83,6 +94,11 @@ mindspore.communication
- **ValueError** - 在后台不可用时抛出。
- **RuntimeError** - 在 `HCCL` 或 `NCCL` 服务不可用时抛出。
样例：
.. note::
.. include:: ops/mindspore.ops.comm_note.rst
.. py:function:: mindspore.communication.get_world_rank_from_group_rank(group, group_rank_id)
由指定通信组中的设备序号获取通信集群中的全局设备序号。
@ -91,7 +107,6 @@ mindspore.communication
- GPU 版本的MindSpore不支持此方法。
- 参数 `group` 不能是 `hccl_world_group`。
- `get_world_rank_from_group_rank` 方法应该在 `init` 方法之后使用。
- 在运行以下示例之前，用户需要预设通信环境变量，请查看mindspore.communication的文档注释。
参数：
- **group** (str) - 传入的通信组名称，通常由 `create_group` 方法创建。
@ -105,6 +120,11 @@ mindspore.communication
- **ValueError** - 参数 `group` 是 `hccl_world_group` 或后台不可用。
- **RuntimeError** - `HCCL` 服务不可用时，或者使用了GPU版本的MindSpore。
样例：
.. note::
.. include:: ops/mindspore.ops.comm_note.rst
.. py:function:: mindspore.communication.get_group_rank_from_world_rank(world_rank_id, group)
由通信集群中的全局设备序号获取指定用户通信组中的rank ID。
@ -113,7 +133,6 @@ mindspore.communication
- GPU 版本的MindSpore不支持此方法。
- 参数 `group` 不能是 `hccl_world_group`。
- `get_group_rank_from_world_rank` 方法应该在 `init` 方法之后使用。
- 在运行以下示例之前，用户需要预设通信环境变量，请查看mindspore.communication的文档注释。
参数：
- **world_rank_id** (`int`) - 通信集群内的全局rank ID。
@ -127,6 +146,11 @@ mindspore.communication
- **ValueError** - 在参数 `group` 是 `hccl_world_group` 或后台不可用时抛出。
- **RuntimeError** - `HCCL` 服务不可用时，或者使用了GPU版本的MindSpore。
样例：
.. note::
.. include:: ops/mindspore.ops.comm_note.rst
.. py:function:: mindspore.communication.create_group(group, rank_ids)
创建用户自定义的通信组实例。
@ -137,7 +161,6 @@ mindspore.communication
- 列表rank_ids内不能有重复数据。
- `create_group` 方法应该在 `init` 方法之后使用。
- 如果没有使用mpirun启动，PyNative模式下仅支持全局单通信组。
- 在运行以下示例之前，用户需要预设通信环境变量，请查看mindspore.communication的文档注释。
参数：
- **group** (str) - 输入用户自定义的通信组实例名称，支持数据类型为str。
@ -148,6 +171,11 @@ mindspore.communication
- **ValueError** - 列表rank_ids的长度小于1，或列表rank_ids内有重复数据，以及后台无效。
- **RuntimeError** - `HCCL` 服务不可用时，或者使用了GPU版本的MindSpore。
样例：
.. note::
.. include:: ops/mindspore.ops.comm_note.rst
.. py:function:: mindspore.communication.get_local_rank(group=GlobalComm.WORLD_COMM_GROUP)
获取指定通信组中当前设备的本地设备序号。
@ -155,7 +183,6 @@ mindspore.communication
.. note::
- GPU 版本的MindSpore不支持此方法。
- `get_local_rank` 方法应该在 `init` 方法之后使用。
- 在运行以下示例之前，用户需要预设通信环境变量，请查看mindspore.communication的文档注释。
参数：
- **group** (`str`) - 通信组名称，通常由 `create_group` 方法创建，否则将使用默认组名称。默认值： `WORLD_COMM_GROUP` 。
@ -168,6 +195,11 @@ mindspore.communication
- **ValueError** - 在后台不可用时抛出。
- **RuntimeError** - `HCCL` 服务不可用时，或者使用了GPU版本的MindSpore。
样例：
.. note::
.. include:: ops/mindspore.ops.comm_note.rst
.. py:function:: mindspore.communication.get_local_rank_size(group=GlobalComm.WORLD_COMM_GROUP)
获取指定通信组的本地设备总数。
@ -175,7 +207,6 @@ mindspore.communication
.. note::
- GPU 版本的MindSpore不支持此方法。
- `get_local_rank_size` 方法应该在 `init` 方法之后使用。
- 在运行以下示例之前，用户需要预设通信环境变量，请查看mindspore.communication的文档注释。
参数：
- **group** (str) - 传入的通信组名称，通常由 `create_group` 方法创建，或默认使用 `WORLD_COMM_GROUP` 。
@ -188,6 +219,11 @@ mindspore.communication
- **ValueError** - 在后台不可用时抛出。
- **RuntimeError** - `HCCL` 服务不可用时，或者使用了GPU版本的MindSpore。
样例：
.. note::
.. include:: ops/mindspore.ops.comm_note.rst
.. py:function:: mindspore.communication.destroy_group(group)
注销用户通信组。

3
docs/api/api_python/mindspore.ops.rst
View File

@ -16,6 +16,7 @@ mindspore.ops
mindspore.ops.adaptive_avg_pool2d
mindspore.ops.adaptive_avg_pool3d
mindspore.ops.adaptive_max_pool1d
mindspore.ops.adaptive_max_pool2d
mindspore.ops.adaptive_max_pool3d
mindspore.ops.avg_pool1d
mindspore.ops.avg_pool2d
@ -359,6 +360,7 @@ Reduction函数
mindspore.ops.matrix_solve
mindspore.ops.mm
mindspore.ops.ger
mindspore.ops.orgqr
mindspore.ops.renorm
mindspore.ops.tensor_dot
@ -413,7 +415,6 @@ Array操作
:nosignatures:
:template: classtemplate.rst
mindspore.ops.adaptive_max_pool2d
mindspore.ops.affine_grid
mindspore.ops.arange
mindspore.ops.argsort

2
docs/api/api_python/nn/mindspore.nn.MaxUnpool1d.rst
View File

@ -10,7 +10,7 @@ mindspore.nn.MaxUnpool1d
.. math::
\begin{array}{ll} \\
H_{out} = (H{in} - 1) \times stride[0] - 2 \times padding[0] + kernel_size[0] \\
H_{out} = (H{in} - 1) \times stride[0] - 2 \times padding[0] + kernel\_size[0] \\
\end{array}
参数：

4
docs/api/api_python/nn/mindspore.nn.MaxUnpool2d.rst
View File

@ -10,8 +10,8 @@ mindspore.nn.MaxUnpool2d
.. math::
\begin{array}{ll} \\
H_{out} = (H{in} - 1) \times stride[0] - 2 \times padding[0] + kernel_size[0] \\
W_{out} = (W{in} - 1) \times stride[1] - 2 \times padding[1] + kernel_size[1] \\
H_{out} = (H{in} - 1) \times stride[0] - 2 \times padding[0] + kernel\_size[0] \\
W_{out} = (W{in} - 1) \times stride[1] - 2 \times padding[1] + kernel\_size[1] \\
\end{array}
参数：

6
docs/api/api_python/nn/mindspore.nn.MaxUnpool3d.rst
View File

@ -10,9 +10,9 @@ mindspore.nn.MaxUnpool3d
.. math::
\begin{array}{ll} \\
D_{out} = (D{in} - 1) \times stride[0] - 2 \times padding[0] + kernel_size[0] \\
H_{out} = (H{in} - 1) \times stride[1] - 2 \times padding[1] + kernel_size[1] \\
W_{out} = (W{in} - 1) \times stride[2] - 2 \times padding[2] + kernel_size[2] \\
D_{out} = (D{in} - 1) \times stride[0] - 2 \times padding[0] + kernel\_size[0] \\
H_{out} = (H{in} - 1) \times stride[1] - 2 \times padding[1] + kernel\_size[1] \\
W_{out} = (W{in} - 1) \times stride[2] - 2 \times padding[2] + kernel\_size[2] \\
\end{array}
参数：

2
docs/api/api_python/ops/mindspore.ops.Custom.rst
View File

@ -5,7 +5,7 @@ mindspore.ops.Custom
`Custom` 算子是MindSpore自定义算子的统一接口。用户可以利用该接口自行定义MindSpore内置算子库尚未包含的算子。
根据输入函数的不用，你可以创建多个自定义算子，并且把它们用在神经网络中。
关于自定义算子的详细说明和介绍，包括参数的正确书写，见 `教程 <https://www.mindspore.cn/tutorials/experts/zh-CN/master/operation/op_custom.html>`_ 。
关于自定义算子的详细说明和介绍，包括参数的正确书写，见 `自定义算子教程 <https://www.mindspore.cn/tutorials/experts/zh-CN/master/operation/op_custom.html>`_ 。
.. warning::
这是一个实验性接口，后续可能删除或修改。

4
docs/api/api_python/ops/mindspore.ops.ExtractVolumePatches.rst
View File

@ -6,9 +6,9 @@ mindspore.ops.ExtractVolumePatches
从输入中提取数据，并将它放入"depth"输出维度中，"depth"为输出的第二维。
参数：
- **kernel_size** (Union[int, tuple[int], list[int]]) - 长度为3或5的int列表。输入每个维度表示滑动窗口大小。必须是：[1, 1, k_d, k_h, k_w]或[k_d, k_h, k_w]。如果k_d = k_h = k_w，则可以输入整数。
- **kernel_size** (Union[int, tuple[int], list[int]]) - 长度为3或5的int列表。输入每个维度表示滑动窗口大小。必须是： :math:`[1, 1, k_d, k_h, k_w]` 或 :math:`[k_d, k_h, k_w]` 。如果 :math:`k_d = k_h = k_w` ，则可以输入整数。
- **strides** (Union[int, tuple[int], list[int]]) - 长度为3或5的int列表。
两个连续色块的中心在输入中的距离。必须是：[1, 1, s_d, s_h, s_w]或[s_d, s_h, s_w]。如果s_d = s_h = s_w，则可以输入整数。
两个连续色块的中心在输入中的距离。必须是： :math:`[1, 1, s_d, s_h, s_w]` 或 :math:`[s_d, s_h, s_w]` 。如果 :math:`s_d = s_h = s_w` ，则可以输入整数。
- **padding** (str) - 要使用的填充算法的类型。可选值有"SAME"和"VALID"。
输入：

2
docs/api/api_python/ops/mindspore.ops.InplaceSub.rst
View File

@ -3,6 +3,6 @@ mindspore.ops.InplaceSub
.. py:class:: mindspore.ops.InplaceSub(indices)
从 `x` 的特定行减去 `input_v` 。即计算 `y` = `x`; y[i,] -= `input_v` 。
从 `x` 的特定行减去 `input_v` 。即计算 :math:`y = x`; :math:`y[i,] -= input\_v` 。
更多参考详见 :func:`mindspore.ops.inplace_sub`。

2
docs/api/api_python/ops/mindspore.ops.ListDiff.rst
View File

@ -7,7 +7,7 @@ mindspore.ops.ListDiff
给定一个列表 `x` 和一个列表 `y`，此操作返回一个列表 `out`，表示在 `x` 中但不在 `y` 中的所有值。
返回列表 `out` 的排序顺序与数字出现在 `x` 中的顺序相同（保留重复项）。此操作还会返回一个列表 `idx` ，
表示每个 `out` 元素在 `x` 中的位置。即： :math:`out[i] = x[idx[i]] for i in [0, 1, ..., len(out) - 1]` 。
表示每个 `out` 元素在 `x` 中的位置。即： :code:`out[i] = x[idx[i]] for i in [0, 1, ..., len(out) - 1]` 。
参数：
- **out_idx** (:class:`mindspore.dtype`，可选) - `idx` 的数据类型，可选值： `mindspore.dtype.int32` 和 `mindspore.dtype.int64` 。默认值： `mindspore.dtype.int32` 。

4
docs/api/api_python/ops/mindspore.ops.Lstsq.rst
View File

@ -5,7 +5,7 @@ mindspore.ops.Lstsq
计算满秩矩阵 `x` :math:`(m \times n)` 与满秩矩阵 `a` :math:`(m \times k)` 的最小二乘问题或最小范数问题的解。
若 :math:`m \geq n` ， `lstsq` 解决最小二乘问题：
若 :math:`m \geq n` ， `Lstsq` 解决最小二乘问题：
.. math::
@ -13,7 +13,7 @@ mindspore.ops.Lstsq
\min_y & \|xy-a\|_2.
\end{array}
若 :math:`m < n` ， `lstsq` 解决最小范数问题：
若 :math:`m < n` ， `Lstsq` 解决最小范数问题：
.. math::

2
docs/api/api_python/ops/mindspore.ops.LuUnpack.rst
View File

@ -26,6 +26,6 @@ mindspore.ops.LuUnpack
- **ValueError** - 若 `LU_data` 的维度小于2。
- **ValueError** - 若 `LU_pivots` 的维度小于1。
- **ValueError** - 若 `LU_pivots` 最后一维的大小不等于 `LU_data` 的最后两维的较小者。
- **ValueError** - 若 `lu_data` 与 `LU_pivots` 的batch维度不匹配。
- **ValueError** - 若 `LU_data` 与 `LU_pivots` 的batch维度不匹配。
- **ValueError** - 在CPU平台上，若 `LU_pivots` 的值不在 :math:`[1, LU\_data.shape[-2]]` 范围内。
- **RuntimeError** - 在Ascend平台上，若 `LU_pivots` 的值不在 :math:`[1, LU\_data.shape[-2]]` 范围内。

2
docs/api/api_python/ops/mindspore.ops.Mod.rst
View File

@ -14,7 +14,7 @@
- 输入数据不支持0。
- 当输出包含的元素个数超过2048时，该算子不能保证双千分之一的精度要求。
- 由于架构的差异，在NPU和CPU上生成的结果可能不一致。
- 如果shape表示为 :math:`(D1、D2...、Dn)` ，则D1\*D2... \*DN<=1000000,n<=8。
- 如果shape表示为 :math:`(D1、D2...、Dn)` ，则 :math:`D1*D2... *DN<=1000000,n<=8` 。
输入：
- **x** (Union[Tensor, numbers.Number, bool]) - 第一个输入是数值型、bool，或数据类型为数值型的Tensor。

2
docs/api/api_python/ops/mindspore.ops.NeighborExchangeV2.rst
View File

@ -5,7 +5,7 @@ mindspore.ops.NeighborExchangeV2
NeighborExchangeV2是一个集合通讯操作。
将数据从本地rank发送到 `send_rank_ids` 中指定的rank，同时从 `recv_rank_ids` 接收数据。请参考 `MindSpore <https://www.mindspore.cn/tutorials/experts/zh-CN/master/parallel/communicate_ops.html#neighborexchangev2>`_ 了解具体的数据是如何在相邻设备间交换的。
将数据从本地rank发送到 `send_rank_ids` 中指定的rank，同时从 `recv_rank_ids` 接收数据。请参考 `分布式集合通信原语 - NeighborExchangeV2 <https://www.mindspore.cn/tutorials/experts/zh-CN/master/parallel/communicate_ops.html#neighborexchangev2>`_ 了解具体的数据是如何在相邻设备间交换的。
.. note::
在运行以下示例之前，用户需要预置环境变量，请在 `NeighborExchangeV2数据交换 <https://www.mindspore.cn/docs/zh-CN/master/api_python/mindspore.ops.html#通信算子>`_ 的官方网站上查看详细信息。

2
docs/api/api_python/ops/mindspore.ops.NextAfter.rst
View File

@ -8,7 +8,7 @@ mindspore.ops.NextAfter
比如有两个数 :math:`a` ， :math:`b` ，数据类型为float32。并且设float32数据类型的可表示值增量为 :math:`eps` 。如果 :math:`a < b` ，那么 :math:`a` 指向 :math:`b` 的下一个可表示值就是 :math:`a+eps` ， :math:`b` 指向 :math:`a` 的下一个可表示值就是 :math:`b-eps` 。
.. math::
out_{i} = nextafter{x1_{i}, x2_{i}}
out_{i} = nextafter({x1_{i}, x2_{i}})
更多详细信息请参见 `A Self Regularized Non-Monotonic Neural Activation Function <https://arxiv.org/abs/1908.08681>`_ 。

2
docs/api/api_python/ops/mindspore.ops.ReduceScatter.rst
View File

@ -3,7 +3,7 @@ mindspore.ops.ReduceScatter
.. py:class:: mindspore.ops.ReduceScatter(op=ReduceOp.SUM, group=GlobalComm.WORLD_COMM_GROUP)
规约并且分发指定通信组中的张量。更多细节请参考 `ReduceScatter <https://www.mindspore.cn/tutorials/experts/zh-CN/master/parallel/communicate_ops.html#reducescatter>`_ 。
规约并且分发指定通信组中的张量。更多细节请参考 `分布式集合通信原语 - ReduceScatter <https://www.mindspore.cn/tutorials/experts/zh-CN/master/parallel/communicate_ops.html#reducescatter>`_ 。
.. note::
在集合的所有过程中，Tensor必须具有相同的shape和格式。

2
docs/api/api_python/ops/mindspore.ops.SegmentMean.rst
View File

@ -5,7 +5,7 @@ mindspore.ops.SegmentMean
计算Tensor的片段均值。
计算一个Tensor，使得 :math:`output_i=\mean_j(input\_x_j)` ，其中求和是在j上，满足：:math:`segment\_ids[j] == i` 。如果给定 ID :math:`i` 的分段的和为空，则有 :math:`output[i] = 0` 。
计算一个Tensor，使得 :math:`output_i=mean_j(input\_x_j)` ，其中求和是在j上，满足：:math:`segment\_ids[j] == i` 。如果给定 ID :math:`i` 的分段的和为空，则有 :math:`output[i] = 0` 。
.. warning::
如果 `input_x` 的数据类型是复数，则无法计算其梯度。

2
docs/api/api_python/ops/mindspore.ops.func_inplace_sub.rst
View File

@ -3,7 +3,7 @@ mindspore.ops.inplace_sub
.. py:function:: mindspore.ops.inplace_sub(x, v, indices)
将 `v` 依照索引 `indices` 从 `x` 中减去。计算 `y` = `x`; y[i,] -= `v`。
将 `v` 依照索引 `indices` 从 `x` 中减去。计算 :math:`y = x`; :math:`y[i,] -= input\_v`。
.. note::
`indices` 只能沿着最高轴进行索引。

2
docs/api/api_python/ops/mindspore.ops.func_max_unpool1d.rst
View File

@ -10,7 +10,7 @@ mindspore.ops.max_unpool1d
.. math::
\begin{array}{ll} \\
H_{out} = (H{in} - 1) \times stride[0] - 2 \times padding[0] + kernel_size[0] \\
H_{out} = (H{in} - 1) \times stride[0] - 2 \times padding[0] + kernel\_size[0] \\
\end{array}
参数：

4
docs/api/api_python/ops/mindspore.ops.func_max_unpool2d.rst
View File

@ -10,8 +10,8 @@ mindspore.ops.max_unpool2d
.. math::
\begin{array}{ll} \\
H_{out} = (H{in} - 1) \times stride[0] - 2 \times padding[0] + kernel_size[0] \\
W_{out} = (W{in} - 1) \times stride[1] - 2 \times padding[1] + kernel_size[1] \\
H_{out} = (H{in} - 1) \times stride[0] - 2 \times padding[0] + kernel\_size[0] \\
W_{out} = (W{in} - 1) \times stride[1] - 2 \times padding[1] + kernel\_size[1] \\
\end{array}
参数：

6
docs/api/api_python/ops/mindspore.ops.func_max_unpool3d.rst
View File

@ -10,9 +10,9 @@ mindspore.ops.max_unpool3d
.. math::
\begin{array}{ll} \\
D_{out} = (D{in} - 1) \times stride[0] - 2 \times padding[0] + kernel_size[0] \\
H_{out} = (H{in} - 1) \times stride[1] - 2 \times padding[1] + kernel_size[1] \\
W_{out} = (W{in} - 1) \times stride[2] - 2 \times padding[2] + kernel_size[2] \\
D_{out} = (D{in} - 1) \times stride[0] - 2 \times padding[0] + kernel\_size[0] \\
H_{out} = (H{in} - 1) \times stride[1] - 2 \times padding[1] + kernel\_size[1] \\
W_{out} = (W{in} - 1) \times stride[2] - 2 \times padding[2] + kernel\_size[2] \\
\end{array}
参数：

3
docs/api/api_python_en/mindspore.ops.rst
View File

@ -16,6 +16,7 @@ Neural Network
mindspore.ops.adaptive_avg_pool2d
mindspore.ops.adaptive_avg_pool3d
mindspore.ops.adaptive_max_pool1d
mindspore.ops.adaptive_max_pool2d
mindspore.ops.adaptive_max_pool3d
mindspore.ops.avg_pool1d
mindspore.ops.avg_pool2d
@ -359,6 +360,7 @@ Linear Algebraic Functions
mindspore.ops.matrix_solve
mindspore.ops.mm
mindspore.ops.ger
mindspore.ops.orgqr
mindspore.ops.renorm
mindspore.ops.tensor_dot
@ -413,7 +415,6 @@ Array Operation
:nosignatures:
:template: classtemplate.rst
mindspore.ops.adaptive_max_pool2d
mindspore.ops.affine_grid
mindspore.ops.arange
mindspore.ops.argsort

2
mindspore/python/mindspore/communication/__init__.py
View File

@ -15,7 +15,7 @@
"""
Collective communication interface.
Note the API in the file needs to preset communication environment variables.
Note that the APIs in the following list need to preset communication environment variables.
For the Ascend devices, users need to prepare the rank table, set rank_id and device_id.
Please see the `Ascend tutorial

123
mindspore/python/mindspore/communication/management.py
View File

@ -96,8 +96,6 @@ def init(backend_name=None):
- The full name of HCCL is Huawei Collective Communication Library.
- The full name of NCCL is NVIDIA Collective Communication Library.
- The full name of MCCL is MindSpore Collective Communication Library.
- The user needs to preset communication environment variables
before running the following example, please see the docstring of the mindspore.communication.
Args:
backend_name (str): Backend, using HCCL/NCCL/MCCL. HCCL should be used for Ascend hardware platforms and
@ -114,6 +112,17 @@ def init(backend_name=None):
``Ascend`` ``GPU``
Examples:
.. note::
Before running the following examples, you need to configure the communication environment variables.
For the Ascend devices, users need to prepare the rank table, set rank_id and device_id.
Please see the `Ascend tutorial
<https://www.mindspore.cn/tutorials/experts/en/master/parallel/train_ascend.html#preparations>`_
for more details.
For the GPU devices, users need to prepare the host file and mpi, please see the `GPU tutorial
<https://www.mindspore.cn/tutorials/experts/en/master/parallel/train_gpu.html#preparation>`_ .
>>> from mindspore.communication import init
>>> init()
"""
@ -169,8 +178,7 @@ def release():
Release distributed resource. e.g. HCCL/NCCL.
Note:
This method should be used after init(). The user needs to preset communication environment variables
before running the following example, please see the docstring of the mindspore.communication.
This method should be used after init().
Raises:
RuntimeError: If failed to release distributed resource.
@ -179,6 +187,17 @@ def release():
``Ascend`` ``GPU``
Examples:
.. note::
Before running the following examples, you need to configure the communication environment variables.
For the Ascend devices, users need to prepare the rank table, set rank_id and device_id.
Please see the `Ascend tutorial
<https://www.mindspore.cn/tutorials/experts/en/master/parallel/train_ascend.html#preparations>`_
for more details.
For the GPU devices, users need to prepare the host file and mpi, please see the `GPU tutorial
<https://www.mindspore.cn/tutorials/experts/en/master/parallel/train_gpu.html#preparation>`_ .
>>> from mindspore.communication import init, release
>>> init()
>>> release()
@ -191,8 +210,7 @@ def get_rank(group=GlobalComm.WORLD_COMM_GROUP):
Get the rank ID for the current device in the specified collective communication group.
Note:
This method should be used after init(). The user needs to preset communication environment variables
before running the following example, please see the docstring of the mindspore.communication.
This method should be used after init().
Args:
group (str): The communication group to work on. Normally, the group should be created by create_group,
@ -210,6 +228,17 @@ def get_rank(group=GlobalComm.WORLD_COMM_GROUP):
``Ascend`` ``GPU``
Examples:
.. note::
Before running the following examples, you need to configure the communication environment variables.
For the Ascend devices, users need to prepare the rank table, set rank_id and device_id.
Please see the `Ascend tutorial
<https://www.mindspore.cn/tutorials/experts/en/master/parallel/train_ascend.html#preparations>`_
for more details.
For the GPU devices, users need to prepare the host file and mpi, please see the `GPU tutorial
<https://www.mindspore.cn/tutorials/experts/en/master/parallel/train_gpu.html#preparation>`_ .
>>> from mindspore.communication import init, get_rank
>>> init()
>>> rank_id = get_rank()
@ -228,8 +257,7 @@ def get_local_rank(group=GlobalComm.WORLD_COMM_GROUP):
Note:
GPU version of MindSpore doesn't support this method.
This method should be used after init(). The user needs to preset communication environment variables
before running the following example, please see the docstring of the mindspore.communication.
This method should be used after init().
Args:
group (str): The communication group to work on. Normally, the group should be created by create_group,
@ -247,6 +275,17 @@ def get_local_rank(group=GlobalComm.WORLD_COMM_GROUP):
``Ascend``
Examples:
.. note::
Before running the following examples, you need to configure the communication environment variables.
For the Ascend devices, users need to prepare the rank table, set rank_id and device_id.
Please see the `Ascend tutorial
<https://www.mindspore.cn/tutorials/experts/en/master/parallel/train_ascend.html#preparations>`_
for more details.
For the GPU devices, users need to prepare the host file and mpi, please see the `GPU tutorial
<https://www.mindspore.cn/tutorials/experts/en/master/parallel/train_gpu.html#preparation>`_ .
>>> import mindspore as ms
>>> from mindspore.communication.management import init, get_rank, get_local_rank
>>> ms.set_context(device_target="Ascend")
@ -268,8 +307,7 @@ def get_group_size(group=GlobalComm.WORLD_COMM_GROUP):
Get the rank size of the specified collective communication group.
Note:
This method should be used after init(). The user needs to preset communication environment variables before
running the following example, please see the docstring of the mindspore.communication.
This method should be used after init().
Args:
group (str): The communication group to work on. Normally, the group should be created by create_group,
@ -287,6 +325,17 @@ def get_group_size(group=GlobalComm.WORLD_COMM_GROUP):
``Ascend`` ``GPU``
Examples:
.. note::
Before running the following examples, you need to configure the communication environment variables.
For the Ascend devices, users need to prepare the rank table, set rank_id and device_id.
Please see the `Ascend tutorial
<https://www.mindspore.cn/tutorials/experts/en/master/parallel/train_ascend.html#preparations>`_
for more details.
For the GPU devices, users need to prepare the host file and mpi, please see the `GPU tutorial
<https://www.mindspore.cn/tutorials/experts/en/master/parallel/train_gpu.html#preparation>`_ .
>>> import mindspore as ms
>>> from mindspore.communication.management import init, get_group_size
>>> ms.set_auto_parallel_context(device_num=8)
@ -307,8 +356,7 @@ def get_local_rank_size(group=GlobalComm.WORLD_COMM_GROUP):
Note:
GPU version of MindSpore doesn't support this method.
This method should be used after init(). The user needs to preset communication environment variables before
running the following example, please see the docstring of the mindspore.communication.
This method should be used after init().
Args:
group (str): The communication group to work on. The group is created by create_group
@ -326,6 +374,17 @@ def get_local_rank_size(group=GlobalComm.WORLD_COMM_GROUP):
``Ascend``
Examples:
.. note::
Before running the following examples, you need to configure the communication environment variables.
For the Ascend devices, users need to prepare the rank table, set rank_id and device_id.
Please see the `Ascend tutorial
<https://www.mindspore.cn/tutorials/experts/en/master/parallel/train_ascend.html#preparations>`_
for more details.
For the GPU devices, users need to prepare the host file and mpi, please see the `GPU tutorial
<https://www.mindspore.cn/tutorials/experts/en/master/parallel/train_gpu.html#preparation>`_ .
>>> import mindspore as ms
>>> from mindspore.communication.management import init, get_local_rank_size
>>> ms.set_context(device_target="Ascend")
@ -349,8 +408,7 @@ def get_world_rank_from_group_rank(group, group_rank_id):
Note:
GPU version of MindSpore doesn't support this method.
The parameter group should not be "hccl_world_group".
This method should be used after init(). The user needs to preset communication environment variables
before running the following example, please see the docstring of the mindspore.communication.
This method should be used after init().
Args:
group (str): The communication group to work on. The group is created by create_group.
@ -368,6 +426,17 @@ def get_world_rank_from_group_rank(group, group_rank_id):
``Ascend``
Examples:
.. note::
Before running the following examples, you need to configure the communication environment variables.
For the Ascend devices, users need to prepare the rank table, set rank_id and device_id.
Please see the `Ascend tutorial
<https://www.mindspore.cn/tutorials/experts/en/master/parallel/train_ascend.html#preparations>`_
for more details.
For the GPU devices, users need to prepare the host file and mpi, please see the `GPU tutorial
<https://www.mindspore.cn/tutorials/experts/en/master/parallel/train_gpu.html#preparation>`_ .
>>> from mindspore import set_context
>>> from mindspore.communication.management import init, create_group, get_world_rank_from_group_rank
>>> set_context(device_target="Ascend")
@ -394,8 +463,6 @@ def get_group_rank_from_world_rank(world_rank_id, group):
GPU version of MindSpore doesn't support this method.
The parameter group should not be "hccl_world_group".
This method should be used after init().
The user needs to preset communication environment variables before running the following example, please see
the docstring of the mindspore.managerment.
Args:
world_rank_id (int): A rank ID in the world communication group.
@ -413,6 +480,17 @@ def get_group_rank_from_world_rank(world_rank_id, group):
``Ascend``
Examples:
.. note::
Before running the following examples, you need to configure the communication environment variables.
For the Ascend devices, users need to prepare the rank table, set rank_id and device_id.
Please see the `Ascend tutorial
<https://www.mindspore.cn/tutorials/experts/en/master/parallel/train_ascend.html#preparations>`_
for more details.
For the GPU devices, users need to prepare the host file and mpi, please see the `GPU tutorial
<https://www.mindspore.cn/tutorials/experts/en/master/parallel/train_gpu.html#preparation>`_ .
>>> from mindspore import set_context
>>> from mindspore.communication.management import init, create_group, get_group_rank_from_world_rank
>>> set_context(device_target="Ascend")
@ -439,8 +517,6 @@ def create_group(group, rank_ids):
The size of rank_ids should be larger than 1, rank_ids should not have duplicate data.
This method should be used after init().
Only support global single communication group in PyNative mode if you do not start with mpirun.
The user needs to preset communication environment variables before running the following example, please see
the docstring of the mindspore.managerment.
Args:
group (str): The name of the communication group to be created.
@ -455,6 +531,17 @@ def create_group(group, rank_ids):
``Ascend``
Examples:
.. note::
Before running the following examples, you need to configure the communication environment variables.
For the Ascend devices, users need to prepare the rank table, set rank_id and device_id.
Please see the `Ascend tutorial
<https://www.mindspore.cn/tutorials/experts/en/master/parallel/train_ascend.html#preparations>`_
for more details.
For the GPU devices, users need to prepare the host file and mpi, please see the `GPU tutorial
<https://www.mindspore.cn/tutorials/experts/en/master/parallel/train_gpu.html#preparation>`_ .
>>> from mindspore import set_context
>>> import mindspore.ops as ops
>>> from mindspore.communication.management import init, create_group

12
mindspore/python/mindspore/nn/layer/pooling.py
View File

@ -1474,7 +1474,7 @@ class MaxUnpool1d(Cell):
.. math::
\begin{array}{ll} \\
H_{out} = (H{in} - 1) \times stride[0] - 2 \times padding[0] + kernel_size[0] \\
H_{out} = (H{in} - 1) \times stride[0] - 2 \times padding[0] + kernel\_size[0] \\
\end{array}
Args:
@ -1554,8 +1554,8 @@ class MaxUnpool2d(Cell):
.. math::
\begin{array}{ll} \\
H_{out} = (H{in} - 1) \times stride[0] - 2 \times padding[0] + kernel_size[0] \\
W_{out} = (W{in} - 1) \times stride[1] - 2 \times padding[1] + kernel_size[1] \\
H_{out} = (H{in} - 1) \times stride[0] - 2 \times padding[0] + kernel\_size[0] \\
W_{out} = (W{in} - 1) \times stride[1] - 2 \times padding[1] + kernel\_size[1] \\
\end{array}
Args:
@ -1644,9 +1644,9 @@ class MaxUnpool3d(Cell):
.. math::
\begin{array}{ll} \\
D_{out} = (D{in} - 1) \times stride[0] - 2 \times padding[0] + kernel_size[0] \\
H_{out} = (H{in} - 1) \times stride[1] - 2 \times padding[1] + kernel_size[1] \\
W_{out} = (W{in} - 1) \times stride[2] - 2 \times padding[2] + kernel_size[2] \\
D_{out} = (D{in} - 1) \times stride[0] - 2 \times padding[0] + kernel\_size[0] \\
H_{out} = (H{in} - 1) \times stride[1] - 2 \times padding[1] + kernel\_size[1] \\
W_{out} = (W{in} - 1) \times stride[2] - 2 \times padding[2] + kernel\_size[2] \\
\end{array}
Args:

4
mindspore/python/mindspore/ops/function/math_func.py
View File

@ -1475,8 +1475,8 @@ def inplace_index_add(var, indices, updates, axis):
def inplace_sub(x, v, indices):
"""
Subtracts `v` into specified rows of `x`. Computes `y` = `x`; y[i,] -= `v`.
r"""
Subtracts `v` into specified rows of `x`. Computes :math:`y = x`; :math:`y[i,] -= input\_v`.
Note:
`indices` refers to the left-most dimension.

12
mindspore/python/mindspore/ops/function/nn_func.py
View File

@ -698,7 +698,7 @@ def max_unpool1d(x, indices, kernel_size, stride=None, padding=0, output_size=No
.. math::
\begin{array}{ll} \\
H_{out} = (H{in} - 1) \times stride[0] - 2 \times padding[0] + kernel_size[0] \\
H_{out} = (H{in} - 1) \times stride[0] - 2 \times padding[0] + kernel\_size[0] \\
\end{array}
Args:
@ -797,8 +797,8 @@ def max_unpool2d(x, indices, kernel_size, stride=None, padding=0, output_size=No
.. math::
\begin{array}{ll} \\
H_{out} = (H{in} - 1) \times stride[0] - 2 \times padding[0] + kernel_size[0] \\
W_{out} = (W{in} - 1) \times stride[1] - 2 \times padding[1] + kernel_size[1] \\
H_{out} = (H{in} - 1) \times stride[0] - 2 \times padding[0] + kernel\_size[0] \\
W_{out} = (W{in} - 1) \times stride[1] - 2 \times padding[1] + kernel\_size[1] \\
\end{array}
Args:
@ -898,9 +898,9 @@ def max_unpool3d(x, indices, kernel_size, stride=None, padding=0, output_size=No
.. math::
\begin{array}{ll} \\
D_{out} = (D{in} - 1) \times stride[0] - 2 \times padding[0] + kernel_size[0] \\
H_{out} = (H{in} - 1) \times stride[1] - 2 \times padding[1] + kernel_size[1] \\
W_{out} = (W{in} - 1) \times stride[2] - 2 \times padding[2] + kernel_size[2] \\
D_{out} = (D{in} - 1) \times stride[0] - 2 \times padding[0] + kernel\_size[0] \\
H_{out} = (H{in} - 1) \times stride[1] - 2 \times padding[1] + kernel\_size[1] \\
W_{out} = (W{in} - 1) \times stride[2] - 2 \times padding[2] + kernel\_size[2] \\
\end{array}
Args:

8
mindspore/python/mindspore/ops/function/sparse_unary_func.py
View File

@ -1553,12 +1553,12 @@ def coo_log1p(x: COOTensor) -> COOTensor:
def csr_round(x: CSRTensor) -> CSRTensor:
"""
r"""
Returns half to even of a csr_tensor element-wise.
.. math::
out_i \approx x_i
out_i \\approx x_i
Args:
x (CSRTensor): The input csr_tensor.
@ -1995,12 +1995,12 @@ def coo_isinf(x: COOTensor) -> COOTensor:
def csr_atanh(x: CSRTensor) -> CSRTensor:
"""
r"""
Computes inverse hyperbolic tangent of the input element-wise.
.. math::
out_i = \tanh^{-1}(x_{i})
out_i = \\tanh^{-1}(x_{i})
.. warning::
This is an experimental prototype that is subject to change and/or deletion.

19
mindspore/python/mindspore/ops/operations/array_ops.py
View File

@ -5190,8 +5190,7 @@ class BatchToSpace(PrimitiveWithInfer):
Each list contains 2 integers.
All values must be not less than 0. crops[i] specifies the crop values for the spatial dimension i, which
corresponds to the input dimension i+2. It is required that
:math:`input\_shape[i+2]*block\_size > crops[i][0]+crops[i][1]`
:math:`input\_shape[i+2]*block\_size > crops[i][0]+crops[i][1]` .
Inputs:
- **input_x** (Tensor) - The input tensor. It must be a 4-D tensor, dimension 0 must be divisible by
@ -6537,7 +6536,7 @@ class ListDiff(Primitive):
is sorted in the same order that the numbers appear in `x` (duplicates are
preserved). This operation also returns a list `idx` that represents the
position of each `out` element in `x`. In other words:
:math:`out[i] = x[idx[i]] for i in [0, 1, ..., len(out) - 1]` .
:code:`out[i] = x[idx[i]] for i in [0, 1, ..., len(out) - 1]` .
Args:
out_idx (:class:`mindspore.dtype`, optional): The dtype of `idx`,
@ -6723,11 +6722,11 @@ class ExtractVolumePatches(Primitive):
Args:
kernel_size (Union[int, tuple[int], list[int]]): A list of ints which's length is 3 or 5.
The size of the sliding window for each dimension of input. Must be: [1, 1, k_d, k_h, k_w] or
[k_d, k_h, k_w]. If k_d = k_h = k_w, you can enter an integer.
The size of the sliding window for each dimension of input. Must be: :math:`[1, 1, k_d, k_h, k_w]` or
:math:`[k_d, k_h, k_w]`. If :math:`k_d = k_h = k_w`, you can enter an integer.
strides (Union[int, tuple[int], list[int]]): A list of ints which's length is 3 or 5.
How far the centers of two consecutive patches are in input. Must be: [1, 1, s_d, s_h, s_w] or
[s_d, s_h, s_w]. If s_d = s_h = s_w, you can enter an integer.
How far the centers of two consecutive patches are in input. Must be: :math:`[1, 1, s_d, s_h, s_w]` or
:math:`[s_d, s_h, s_w]`. If :math:`s_d = s_h = s_w`, you can enter an integer.
padding (str): A string from: "SAME", "VALID". The type of padding algorithm to use.
Inputs:
@ -6859,7 +6858,7 @@ class Lstsq(Primitive):
Computes the solutions of the least squares and minimum norm problems of full-rank
matrix `x` of size :math:`(m \times n)` and matrix `a` of size :math:`(m \times k)`.
If :math:`m \geq n`, `lstsq` solves the least-squares problem:
If :math:`m \geq n`, `Lstsq` solves the least-squares problem:
.. math::
@ -6867,7 +6866,7 @@ class Lstsq(Primitive):
\min_y & \|xy-a\|_2.
\end{array}
If :math:`m < n`, `lstsq` solves the least-norm problem:
If :math:`m < n`, `Lstsq` solves the least-norm problem:
.. math::
@ -7680,7 +7679,7 @@ class SegmentMean(Primitive):
r"""
Computes the mean along segments of a tensor.
Computes a tensor such that :math:`output_i = \mean_j input\_x_j` where mean is over :math:`j` such that
Computes a tensor such that :math:`output_i = mean_j(input\_x_j)` where mean is over :math:`j` such that
:math:`segment\_ids[j] == i`. If the mean is empty for a given segment ID :math:`i`, :math:`output[i] = 0`.
.. warning::

4
mindspore/python/mindspore/ops/operations/comm_ops.py
View File

@ -409,7 +409,7 @@ class _HostAllGather(PrimitiveWithInfer):
class ReduceScatter(Primitive):
r"""
Reduces and scatters tensors from the specified communication group.
For more details about it, please refer to `ReduceScatter \
For more details about it, please refer to `Distributed Set Communication Primitives - ReduceScatter \
<https://www.mindspore.cn/tutorials/experts/en/master/parallel/communicate_ops.html#reducescatter>`_ .
Note:
@ -884,7 +884,7 @@ class NeighborExchangeV2(Primitive):
NeighborExchangeV2 sends data from the local rank to ranks in the `send_rank_ids`,
as while receive data from `recv_rank_ids`. Please refer to
`NeighborExchangeV2 data exchange rules \
`Distributed Set Communication Primitives - NeighborExchangeV2 \
<https://www.mindspore.cn/tutorials/experts/zh-CN/master/parallel/communicate_ops.html#neighborexchangev2>`_
to learn about how the data is exchanged between neighborhood devices.

3
mindspore/python/mindspore/ops/operations/custom_ops.py
View File

@ -149,6 +149,9 @@ class Custom(ops.PrimitiveWithInfer):
primitives. You can construct a `Custom` object with a predefined function, which describes the computation
logic of a user defined operator. You can also construct another `Custom` object with another predefined
function if needed. Then these `Custom` objects can be directly used in neural networks.
Detailed description and introduction of user-defined operators, including correct writing of parameters,
please refer to `Custom Operators Tutorial
<https://www.mindspore.cn/tutorials/experts/en/master/operation/op_custom.html>`_ .
.. warning::
This is an experimental prototype that is subject to change.

8
mindspore/python/mindspore/ops/operations/math_ops.py
View File

@ -1996,8 +1996,8 @@ class InplaceIndexAdd(Primitive):
class InplaceSub(Primitive):
"""
Subtracts `v` into specified rows of `x`. Computes `y` = `x`; y[i,] -= `v`.
r"""
Subtracts `v` into specified rows of `x`. Computes :math:`y = x`; :math:`y[i,] -= input\_v`.
Refer to :func:`mindspore.ops.inplace_sub` for more details.
@ -3367,7 +3367,7 @@ class Mod(_MathBinaryOp):
- When the elements of input exceed 2048, the accuracy of operator cannot guarantee the requirement of
double thousandths in the mini form.
- Due to different architectures, the calculation results of this operator on NPU and CPU may be inconsistent.
- If shape is expressed as (D1,D2... ,Dn), then D1\*D2... \*DN<=1000000,n<=8.
- If shape is expressed as :math:`(D1,D2... ,Dn)`, then :math:`D1*D2... *DN<=1000000,n<=8`.
Inputs:
- **x** (Union[Tensor, numbers.Number, bool]) - The first input is a number, a bool
@ -7277,7 +7277,7 @@ class NextAfter(Primitive):
.. math::
out_{i} = nextafter{x1_{i}, x2_{i}}
out_{i} = nextafter({x1_{i}, x2_{i}})
Inputs:
- **x1** (Tensor) - The shape of tensor is