mindspore 2.4.10__cp39-cp39-manylinux1_x86_64.whl → 2.5.0__cp39-cp39-manylinux1_x86_64.whl

mindspore/ops/operations/comm_ops.py

@@ -18,6 +18,7 @@
 from __future__ import absolute_import
 from __future__ import division
 
+import os
 from mindspore.common import Tensor
 from mindspore import _checkparam as validator
 from mindspore.communication.management import get_rank, get_group_size, GlobalComm, _get_group, _host_distribute
@@ -54,7 +55,7 @@ class ReduceOp:
             For Ascend/GPU/CPU devices, it is recommended to use the msrun startup method
             without any third-party or configuration file dependencies.
             Please see the `msrun start up
-            <https://www.mindspore.cn/docs/zh-CN/master/model_train/parallel/msrun_launcher.html>`_
+            <https://www.mindspore.cn/docs/en/master/model_train/parallel/msrun_launcher.html>`_
             for more details.
 
             This example should be run with multiple devices.
@@ -100,7 +101,9 @@ def check_collective_target_dtype(data_name, data_dtype, prim_name):
 def check_hcom_group_valid(group, prim_name=None):
     """Check if hcom group is valid."""
     msg_prefix = f"For '{prim_name}', the" if prim_name else "The"
-    if not _host_distribute() and context.get_context("mode") == context.PYNATIVE_MODE and \
+    sim_level = os.getenv("MS_SIMULATION_LEVEL")
+    no_sim = (sim_level is None or sim_level.strip() == '')
+    if no_sim and (not _host_distribute()) and context.get_context("mode") == context.PYNATIVE_MODE and \
             group != GlobalComm.WORLD_COMM_GROUP:
         raise RuntimeError(f"{msg_prefix} 'group' only support 'hccl_world_group' in pynative mode, but got "
                            f"'group': {group}. Please start by using mpi-run.")
@@ -141,7 +144,7 @@ class AllReduce(Primitive):
             For Ascend/GPU/CPU devices, it is recommended to use the msrun startup method
             without any third-party or configuration file dependencies.
             Please see the `msrun start up
-            <https://www.mindspore.cn/docs/zh-CN/master/model_train/parallel/msrun_launcher.html>`_
+            <https://www.mindspore.cn/docs/en/master/model_train/parallel/msrun_launcher.html>`_
             for more details.
 
             This example should be run with 2 devices.
@@ -231,7 +234,7 @@ class Reduce(PrimitiveWithInfer):
             For Ascend/GPU/CPU devices, it is recommended to use the msrun startup method without any third-party
             or configuration file dependencies.
             Please see the `msrun start up
-            <https://www.mindspore.cn/docs/zh-CN/master/model_train/parallel/msrun_launcher.html>`_
+            <https://www.mindspore.cn/docs/en/master/model_train/parallel/msrun_launcher.html>`_
             for more details.
 
             This example should be run with 4 devices.
@@ -315,7 +318,7 @@ class AllGather(PrimitiveWithInfer):
             For Ascend/GPU/CPU devices, it is recommended to use the msrun startup method
             without any third-party or configuration file dependencies.
             Please see the `msrun start up
-            <https://www.mindspore.cn/docs/zh-CN/master/model_train/parallel/msrun_launcher.html>`_
+            <https://www.mindspore.cn/docs/en/master/model_train/parallel/msrun_launcher.html>`_
             for more details.
 
             This example should be run with 2 devices.
@@ -538,7 +541,7 @@ class ReduceScatter(Primitive):
             For Ascend/GPU/CPU devices, it is recommended to use the msrun startup method
             without any third-party or configuration file dependencies.
             Please see the `msrun start up
-            <https://www.mindspore.cn/docs/zh-CN/master/model_train/parallel/msrun_launcher.html>`_
+            <https://www.mindspore.cn/docs/en/master/model_train/parallel/msrun_launcher.html>`_
             for more details.
 
             This example should be run with 2 devices.
@@ -676,7 +679,7 @@ class Broadcast(PrimitiveWithInfer):
             For Ascend/GPU/CPU devices, it is recommended to use the msrun startup method
             without any third-party or configuration file dependencies.
             Please see the `msrun start up
-            <https://www.mindspore.cn/docs/zh-CN/master/model_train/parallel/msrun_launcher.html>`_
+            <https://www.mindspore.cn/docs/en/master/model_train/parallel/msrun_launcher.html>`_
             for more details.
 
             This example should be run with 2 devices.
@@ -906,7 +909,7 @@ class AlltoAll(PrimitiveWithInfer):
             For Ascend/GPU/CPU devices, it is recommended to use the msrun startup method
             without any third-party or configuration file dependencies.
             Please see the `msrun start up
-            <https://www.mindspore.cn/docs/zh-CN/master/model_train/parallel/msrun_launcher.html>`_
+            <https://www.mindspore.cn/docs/en/master/model_train/parallel/msrun_launcher.html>`_
             for more details.
 
             This example should be run with 8 devices.
@@ -1028,7 +1031,7 @@ class NeighborExchangeV2(Primitive):
             For Ascend/GPU/CPU devices, it is recommended to use the msrun startup method
             without any third-party or configuration file dependencies.
             Please see the `msrun start up
-            <https://www.mindspore.cn/docs/zh-CN/master/model_train/parallel/msrun_launcher.html>`_
+            <https://www.mindspore.cn/docs/en/master/model_train/parallel/msrun_launcher.html>`_
             for more details.
 
             This example should be run with 2 devices.
@@ -1145,7 +1148,7 @@ class CollectiveScatter(Primitive):
             For Ascend/GPU/CPU devices, it is recommended to use the msrun startup method
             without any third-party or configuration file dependencies.
             Please see the `msrun start up
-            <https://www.mindspore.cn/docs/zh-CN/master/model_train/parallel/msrun_launcher.html>`_
+            <https://www.mindspore.cn/docs/en/master/model_train/parallel/msrun_launcher.html>`_
             for more details.
 
             This example should be run with 2 devices.
@@ -1230,7 +1233,7 @@ class CollectiveGather(Primitive):
             For Ascend/GPU/CPU devices, it is recommended to use the msrun startup method
             without any third-party or configuration file dependencies.
             Please see the `msrun start up
-            <https://www.mindspore.cn/docs/zh-CN/master/model_train/parallel/msrun_launcher.html>`_
+            <https://www.mindspore.cn/docs/en/master/model_train/parallel/msrun_launcher.html>`_
             for more details.
 
             This example should be run with 4 devices.
@@ -1306,7 +1309,7 @@ class Barrier(PrimitiveWithInfer):
             For Ascend/GPU/CPU devices, it is recommended to use the msrun startup method
             without any third-party or configuration file dependencies.
             Please see the `msrun start up
-            <https://www.mindspore.cn/docs/zh-CN/master/model_train/parallel/msrun_launcher.html>`_
+            <https://www.mindspore.cn/docs/en/master/model_train/parallel/msrun_launcher.html>`_
             for more details.
 
             This example should be run with 2 devices.
@@ -1380,7 +1383,7 @@ class Send(PrimitiveWithInfer):
             For Ascend/GPU/CPU devices, it is recommended to use the msrun startup method
             without any third-party or configuration file dependencies.
             Please see the `msrun start up
-            <https://www.mindspore.cn/docs/zh-CN/master/model_train/parallel/msrun_launcher.html>`_
+            <https://www.mindspore.cn/docs/en/master/model_train/parallel/msrun_launcher.html>`_
             for more details.
 
             This example should be run with 2 devices.
@@ -1464,7 +1467,7 @@ class Receive(PrimitiveWithInfer):
             For Ascend/GPU/CPU devices, it is recommended to use the msrun startup method
             without any third-party or configuration file dependencies.
             Please see the `msrun start up
-            <https://www.mindspore.cn/docs/zh-CN/master/model_train/parallel/msrun_launcher.html>`_
+            <https://www.mindspore.cn/docs/en/master/model_train/parallel/msrun_launcher.html>`_
             for more details.
 
             This example should be run with 2 devices.
@@ -1842,7 +1845,7 @@ class BatchISendIRecv(PrimitiveWithInfer):
             without any third-party or configuration file dependencies.
 
             Please see the `msrun start up
-            <https://www.mindspore.cn/docs/zh-CN/master/model_train/parallel/msrun_launcher.html>`_
+            <https://www.mindspore.cn/docs/en/master/model_train/parallel/msrun_launcher.html>`_
             for more details.
 
             This example should be run with 2 devices.
@@ -1955,7 +1958,7 @@ class AlltoAllV(PrimitiveWithInfer):
             without any third-party or configuration file dependencies.
 
             Please see the `msrun start up
-            <https://www.mindspore.cn/docs/zh-CN/master/model_train/parallel/msrun_launcher.html>`_
+            <https://www.mindspore.cn/docs/en/master/model_train/parallel/msrun_launcher.html>`_
             for more details.
 
             This example should be run with 2 devices.

mindspore/ops/operations/custom_ops.py

@@ -223,8 +223,6 @@ class Custom(ops.PrimitiveWithInfer):
     .. note::
         The supported platforms are determined by the input `func_type`. The supported platforms are as follows:
 
-        - "hybrid": supports ["GPU", "CPU"].
-        - "akg": supports ["GPU", "CPU"].
         - "aot": supports ["GPU", "CPU", "Ascend"].
         - "pyfunc": supports ["CPU"].
         - "julia": supports ["CPU"].
@@ -233,11 +231,7 @@ class Custom(ops.PrimitiveWithInfer):
         func (Union[function, str]):
 
             - function: If func is of function type, then func should be a Python function which describes the
-              computation logic of a user defined operator. The function can be one of the following:
-
-              1. A AKG operator implementation function, which can use ir builder/tvm compute/hybrid grammar.
-              2. A pure python function
-              3. An kernel decorated function written by the Hybrid DSL.
+              computation logic of a user defined operator.
 
             - str: If func is of str type, then str should be a path of file along with a function name.
               This could be used when func_type is "aot" or "julia".
@@ -251,13 +245,11 @@ class Custom(ops.PrimitiveWithInfer):
 
                  - "xxx.so" file generation:
 
-                   1) GPU Platform: Given user defined "xxx.cu" file (ex. "{path}/add.cu"),
-                   use nvcc command to compile
-                   it.(ex. :code:`nvcc --shared -Xcompiler -fPIC -o add.so add.cu`)
+                   1) GPU Platform: Given user defined "xxx.cu" file (ex. "{path}/add.cu"), use nvcc command to compile
+                   it.(ex. "nvcc --shared -Xcompiler -fPIC -o add.so add.cu")
 
-                   2) CPU Platform: Given user defined "xxx.cc" file (ex. "{path}/add.cc"),
-                   use g++/gcc command to
-                   compile it.(ex. :code:`g++ --shared -fPIC -o add.so add.cc`)
+                   2) CPU Platform: Given user defined "xxx.cc" file (ex. "{path}/add.cc"), use g++/gcc command to
+                   compile it.(ex. "g++ --shared -fPIC  -o add.so add.cc")
 
                  - Define a "xxx.cc"/"xxx.cu" file:
 
@@ -356,7 +348,7 @@ class Custom(ops.PrimitiveWithInfer):
 
         func_type (str): The implementation type of `func`, should be one of
 
-            [ ``"hybrid"`` , ``"akg"`` , ``"aot"`` , ``"pyfunc"`` , ``"julia"`` ].
+            [ ``"aot"`` , ``"pyfunc"`` , ``"julia"`` ].
 
         bprop (function): The back propagation function of `func`. Default: ``None`` .
         reg_info (Union[str, dict, list, tuple]): Represents the registration information(reg info) of `func` with
@@ -400,50 +392,14 @@ class Custom(ops.PrimitiveWithInfer):
         >>> input_x = Tensor(np.ones([16, 16]).astype(np.float32))
         >>> input_y = Tensor(np.ones([16, 16]).astype(np.float32))
         >>>
-        >>> # Example, func_type = "hybrid"
-        >>> # This is the default func_type in Custom,
-        >>> # and both out_shape and out_dtype can be None(default value).
-        >>> # In this case, the input func must be a function written in the Hybrid DSL
-        >>> # and decorated by @kernel.
-        >>> @kernel
-        ... def add_script(a, b):
-        ...     c = output_tensor(a.shape, a.dtype)
-        ...     for i0 in range(a.shape[0]):
-        ...         for i1 in range(a.shape[1]):
-        ...             c[i0, i1] = a[i0, i1] + b[i0, i1]
-        ...     return c
-        >>>
-        >>> test_op_hybrid = ops.Custom(add_script)
-        >>> output = test_op_hybrid(input_x, input_y)
-        >>> # the result will be a 16 * 16 tensor with all elements 2
-        >>> print(output.shape)
-        (16, 16)
-        >>> # Example, func_type = "aot"
-        >>> def test_aot(x, y, out_shapes, out_types):
-        ...     program = ops.Custom("./reorganize.so:CustomReorganize", out_shapes, out_types, "aot")
-        ...     out = program(x, y)
-        ...     return out
-        >>>
         >>> # Example, func_type = "pyfunc"
-        >>> def func_multi_output(x1, x2):
-        ...     return (x1 + x2), (x1 - x2)
+        >>> def func_pyfunc(x1, x2):
+        ...     return x1 + x2
         >>>
-        >>> test_pyfunc = ops.Custom(func_multi_output, lambda x, _: (x, x), lambda x, _: (x, x), "pyfunc")
+        >>> test_pyfunc = ops.Custom(func_pyfunc, lambda x, _: x, lambda x, _: x, "pyfunc")
         >>> output = test_pyfunc(input_x, input_y)
-        >>>
-        >>> # Example, func_type = "julia"
-        >>> # julia code:
-        >>> # add.jl
-        >>> # module Add
-        >>> # function add(x, y, z)
-        >>> #   z .= x + y
-        >>> #   return z
-        >>> # end
-        >>> # end
-        >>> def test_julia(x, y, out_shapes, out_types):
-        ...     program = ops.Custom("./add.jl:Add:add", out_shapes, out_types, "julia")
-        ...     out = program(x, y)
-        ...     return out
+        >>> print(output.shape)
+        (16, 16)
     """
 
     registered_func = {}

mindspore/ops/operations/debug_ops.py

@@ -13,23 +13,24 @@
 # limitations under the License.
 # ============================================================================
 """debug_ops"""
-import os
 import stat
+from pathlib import Path
 
 import numpy as np
 from mindspore import log as logger
 from mindspore._c_expression import security, HookType
 from mindspore._c_expression import Tensor as Tensor_
+from mindspore._c_expression import _tensordump_process_file
 from mindspore import _checkparam as validator
 from mindspore.common import dtype as mstype
 from mindspore.common.parameter import Parameter
 from mindspore.common.tensor import Tensor
 from mindspore.ops.primitive import prim_attr_register, Primitive, PrimitiveWithInfer
 from mindspore._checkparam import check_hook_fn
+from mindspore.ops import operations as P
 
 
 SUMMARY_TENSOR_CACHE = []
-TENSORDUMP_ID = 0
 
 
 def _cache_summary_data(op_name, define_name, tensor):
@@ -64,8 +65,8 @@ class ScalarSummary(Primitive):
     which specify the directory of the summary file. The summary file can
     be loaded and shown by MindInsight, see `MindInsight documents <https://www.mindspore.cn/
     mindinsight/docs/en/master/index.html>`_ for details.
-    In Ascend platform with graph mode, can set environment variables `MS_DUMP_SLICE_SIZE` and `MS_DUMP_WAIT_TIME`
-    to solve operator execution failure when calling this operator intensively.
+    In Ascend platform with graph mode, the environment variables `MS_DUMP_SLICE_SIZE` and `MS_DUMP_WAIT_TIME`
+    can be set to solve operator execution failure when calling this operator intensively.
 
     Inputs:
         - **name** (str) - The name of the input variable, it must not be an empty string.
@@ -124,8 +125,8 @@ class ImageSummary(Primitive):
     SummaryRecord or SummaryCollector, which specify the directory of the summary file. The summary file can
     be loaded and shown by MindInsight, see `MindInsight documents <https://www.mindspore.cn/
     mindinsight/docs/en/master/index.html>`_ for details.
-    In Ascend platform with graph mode, can set environment variables `MS_DUMP_SLICE_SIZE` and `MS_DUMP_WAIT_TIME`
-    to solve operator execution failure when calling this operator intensively.
+    In Ascend platform with graph mode, the environment variables `MS_DUMP_SLICE_SIZE` and `MS_DUMP_WAIT_TIME`
+    can be set to solve operator execution failure when calling this operator intensively.
 
     Inputs:
         - **name** (str) - The name of the input variable, it must not be an empty string.
@@ -177,8 +178,8 @@ class TensorSummary(Primitive):
     or SummaryCollector, which specify the directory of the summary file. The summary file can
     be loaded and shown by MindInsight, see `MindInsight documents <https://www.mindspore.cn/
     mindinsight/docs/en/master/index.html>`_ for details.
-    In Ascend platform with graph mode, can set environment variables `MS_DUMP_SLICE_SIZE` and `MS_DUMP_WAIT_TIME`
-    to solve operator execution failure when calling this operator intensively.
+    In Ascend platform with graph mode, the environment variables `MS_DUMP_SLICE_SIZE` and `MS_DUMP_WAIT_TIME`
+    can be set to solve operator execution failure when calling this operator intensively.
 
     Inputs:
         - **name** (str) - The name of the input variable.
@@ -235,15 +236,15 @@ class TensorDump(Primitive):
     Save the Tensor as an npy file in numpy format.
 
     .. warning::
-        - If a large amount of data is stored within a short period, it may lead to memory overflow on the device side.
-          Consider slicing the data to reduce the data scale.
-        - Since data saving is processed asynchronously, when the amount of data is too large or the main process exits
-          too quickly, data loss may occur. You need to actively control the destruction time of the main process,
-          such as using sleep.
+        - The parameter input_output will no longer support the value 'all'.
+
+    .. note::
+        In Ascend platform with graph mode, the environment variables `MS_DUMP_SLICE_SIZE` and `MS_DUMP_WAIT_TIME`
+        can be set to solve operator execution failure when outputting big tensor or outputting tensor intensively.
 
     Args:
         input_output (str, optional): Used to control Tensordump behavior.
-            Available value is one of ['in', 'out', 'all']. Default value is ``out``.
+            Available value is one of ['in', 'out']. Default value is ``out``.
 
             In case of OpA --> RedistributionOps --> OpB,
             The dump data of OpA's output is not equal to OpB's input (Due to the redistribution operators).
@@ -253,21 +254,19 @@ class TensorDump(Primitive):
             Different requirements of saving dump data can be achieved by configuring parameter input_output:
 
             - If the input_output is 'out', the dump data contains only OpA's output slice.
-            - If the input_output is 'all', the dump data contains both OpA's output slice and OpB's input slice.
             - If the input_output is 'in', the dump data contains only OpB's input slice.
 
-            For input_output is 'all' or 'in', the input slice npy file format is:
-            id_fileName_cNodeID_dumpMode_rankID.npy.
+            For input_output is 'in', the input slice npy file format is:
+            fileName_dumpMode_dtype_id.npy.
 
-            For input_output is 'out' or 'all' the output slice npy file format is:
-            id_fileName.npy.
+            For input_output is 'out', the output slice npy file format is:
+            fileName_dtype_id.npy.
 
-            - id: An auto increment ID.
             - fileName: Value of the parameter file
               (if parameter file_name is a user-specified path, the value of fileName is the last level of the path).
-            - cNodeID: The node ID of the Tensordump node in the step_parallel_end.ir file.
             - dumpMode: Value of the parameter input_output.
-            - rankID: Logical device id.
+            - dtype: The original data type. Data of type bfloat16 stored in the .npy file will be converted to float32.
+            - id: An auto increment ID.
 
     Inputs:
         - **file** (str) - The path of the file to be saved.
@@ -285,7 +284,8 @@ class TensorDump(Primitive):
         >>> import mindspore as ms
         >>> import time
         >>> from mindspore import nn, Tensor, ops
-        >>> ms.set_context(mode=ms.GRAPH_MODE, device_target="Ascend")
+        >>> ms.set_context(mode=ms.GRAPH_MODE)
+        >>> ms.set_device(device_target="Ascend")
         >>> class Net(nn.Cell):
         ...     def __init__(self):
         ...         super(Net, self).__init__()
@@ -305,7 +305,7 @@ class TensorDump(Primitive):
         >>> net = Net()
         >>> out = net(input_x)
         >>> time.sleep(0.5)
-        >>> add = np.load('0_add.npy')
+        >>> add = np.load('add_float32_0.npy')
         >>> print(add)
         [[2. 3. 4. 5.]
          [6. 7. 8. 9.]]
@@ -318,25 +318,26 @@ class TensorDump(Primitive):
         self.add_prim_attr("side_effect_io", True)
         self.add_prim_attr("channel_name", "ms_tensor_dump")
 
+    def _save_file(self, file, data):
+        file = Path(file)
+        if file.exists():
+            file.chmod(stat.S_IWUSR)
+        np.save(file, data)
+        file.chmod(stat.S_IRUSR)
+
     def __call__(self, file, input_x):
         validator.check_value_type('file', file, [str], self.__class__.__name__)
         if not file:
             raise ValueError("For 'TensorDump', the input argument[file] cannot be an empty string.")
         validator.check_value_type('input_x', input_x, [Tensor], self.__class__.__name__)
-        global TENSORDUMP_ID
-        npy_suffix = ".npy"
-        directory, filename = os.path.split(file)
-        if directory and not os.path.exists(directory):
-            os.makedirs(directory, mode=0o700, exist_ok=True)
-        new_filename = f"{TENSORDUMP_ID}_{filename}"
-        new_file = os.path.join(directory, new_filename)
-        if not new_file.endswith(npy_suffix):
-            new_file += npy_suffix
-        if os.path.exists(new_file):
-            os.chmod(new_file, stat.S_IWUSR)
-        np.save(new_file, input_x.asnumpy())
-        os.chmod(new_file, stat.S_IRUSR)
-        TENSORDUMP_ID += 1
+
+        dtype = input_x.dtype
+        file = _tensordump_process_file(file, str(dtype))
+        if not file:
+            return
+        if dtype == mstype.bfloat16:
+            input_x = P.Cast()(input_x, mstype.float32)
+        self._save_file(file, input_x.asnumpy())
 
 
 class HistogramSummary(Primitive):
@@ -345,8 +346,8 @@ class HistogramSummary(Primitive):
     It must be used with SummaryRecord or SummaryCollector, which specify the directory of the summary file.
     The summary file can be loaded and shown by MindInsight, see `MindInsight documents <https://www.mindspore.cn/
     mindinsight/docs/en/master/index.html>`_ for details.
-    In Ascend platform with graph mode, can set environment variables `MS_DUMP_SLICE_SIZE` and `MS_DUMP_WAIT_TIME`
-    to solve operator execution failure when calling this operator intensively.
+    In Ascend platform with graph mode, the environment variables `MS_DUMP_SLICE_SIZE` and `MS_DUMP_WAIT_TIME`
+    can be set to solve operator execution failure when calling this operator intensively.
 
     Inputs:
         - **name** (str) - The name of the input variable.
@@ -398,7 +399,7 @@ class HistogramSummary(Primitive):
         _cache_summary_data(self.name, args[0], args[1])
 
 
-class InsertGradientOf(PrimitiveWithInfer):
+class InsertGradientOf(Primitive):
     """
     Attaches callback to the graph node that will be invoked on the node's gradient.
 
@@ -464,12 +465,6 @@ class InsertGradientOf(PrimitiveWithInfer):
         self.add_prim_attr('side_effect_backprop', True)
         self.f = f
 
-    def infer_shape(self, x_shape):
-        return x_shape
-
-    def infer_dtype(self, x_type):
-        return x_type
-
 
 class HookBackward(PrimitiveWithInfer):
     """

mindspore/ops/operations/inner_ops.py

@@ -302,8 +302,8 @@ class FusedWeightScaleApplyMomentum(PrimitiveWithInfer):
           Default: ``0.0`` .
         - **loss_scale** (Tensor) - The loss scale value, must be a scalar tensor with float data type.
           Default: ``1.0`` .
-        - **variable** (Parameter) - Weights to be updated. data type must be float.
-        - **accumulation** (Parameter) - Accumulated gradient value by moment weight.
+        - **variable** (Union[Parameter, Tensor]) - Weights to be updated. data type must be float.
+        - **accumulation** (Union[Parameter, Tensor]) - Accumulated gradient value by moment weight.
           Has the same data type with `variable`.
         - **learning_rate** (Union[Number, Tensor]) - The learning rate value, must be a float number or
           a scalar tensor with float data type.
@@ -425,7 +425,8 @@ class FusedCastAdamWeightDecay(PrimitiveWithInfer):
         ...     def construct(self, lr, beta1, beta2, epsilon, decay, grad, norm):
         ...         out = self.opt(self.var, self.m, self.v, lr, beta1, beta2, epsilon, decay, grad, norm)
         ...         return out
-        >>> ms.set_context(mode=ms.GRAPH_MODE, device_target="CPU")
+        >>> ms.set_context(mode=ms.GRAPH_MODE)
+        >>> ms.set_device(device_target="CPU")
         >>> net = Net()
         >>> gradient = Tensor(np.ones([2, 2]), mstype.float16)
         >>> output = net(0.001, 0.9, 0.999, 1e-8, 0.0, gradient, 1.0)
@@ -560,7 +561,8 @@ class FusedAdaFactor(PrimitiveWithInfer):
         ...         out = self.opt(epsilon, clip_threshold, beta1, beta2, weight_decay, lr, grad, self.param,
         ...                        self.exp_avg, self.exp_avg_sq_row, self.exp_avg_sq_col, self.exp_avg_sq)
         ...         return out
-        >>> ms.set_context(mode=ms.GRAPH_MODE, device_target="CPU")
+        >>> ms.set_context(mode=ms.GRAPH_MODE)
+        >>> ms.set_device(device_target="CPU")
         >>> net = Net()
         >>> gradient = Tensor(np.ones(param_shape), mstype.float32)
         >>> output = net((1e-30, 1e-3), 1.0, 0.9, 0.8, 1e-2, 0.03, gradient)

mindspore/ops/operations/linalg_ops.py

@@ -51,9 +51,10 @@ class Svd(Primitive):
 
     Examples:
         >>> import numpy as np
-        >>> from mindspore import Tensor, set_context
+        >>> import mindspore as ms
+        >>> from mindspore import Tensor
         >>> from mindspore import ops
-        >>> set_context(device_target="CPU")
+        >>> ms.set_device(device_target="CPU")
         >>> svd = ops.Svd(full_matrices=True, compute_uv=True)
         >>> a = Tensor(np.array([[1, 2], [-4, -5], [2, 1]]).astype(np.float32))
         >>> s, u, v = svd(a)