mindspore 2.3.0__cp39-none-any.whl → 2.3.0rc2__cp39-none-any.whl

mindspore/communication/management.py

@@ -21,13 +21,12 @@ from mindspore.communication._comm_helper import Backend, _get_rank_helper, _get
     _get_world_rank_from_group_rank_helper, _get_group_rank_from_world_rank_helper, \
     _create_group_helper, _destroy_group_helper, HCCL_WORLD_COMM_GROUP, NCCL_WORLD_COMM_GROUP, \
     MCCL_WORLD_COMM_GROUP, DEVICE_TO_BACKEND, _get_local_rank_helper, _get_local_size_helper, GlobalComm, \
-    _check_mpi_envs, _set_elegant_exit_handle, _get_group_ranks
+    _check_mpi_envs, _set_elegant_exit_handle
 from mindspore._c_expression import init_hccl, finalize_hccl, init_cluster, MSContext, ms_ctx_param
-from mindspore.hal.device import is_initialized
 
 __all__ = ["init", "release", "get_rank", "get_local_rank", "get_group_size",
            "get_local_rank_size", "get_world_rank_from_group_rank",
-           "get_group_rank_from_world_rank", "create_group", "destroy_group", "get_process_group_ranks",
+           "get_group_rank_from_world_rank", "create_group", "destroy_group",
            "HCCL_WORLD_COMM_GROUP", "NCCL_WORLD_COMM_GROUP", "MCCL_WORLD_COMM_GROUP"]
 
 DEFAULT_WORLD_COMM_GROUP = HCCL_WORLD_COMM_GROUP
@@ -137,12 +136,17 @@ def init(backend_name=None):
         .. note::
             Before running the following examples, you need to configure the communication environment variables.
 
-            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/tutorials/experts/zh-CN/master/parallel/msrun_launcher.html>`_
+            For the Ascend devices, users need to prepare the rank table, set rank_id and device_id.
+            Please see the `rank table Startup
+            <https://www.mindspore.cn/tutorials/experts/en/master/parallel/rank_table.html>`_
             for more details.
 
+            For the GPU devices, users need to prepare the host file and mpi, please see the `mpirun Startup
+            <https://www.mindspore.cn/tutorials/experts/en/master/parallel/mpirun.html>`_ .
+
+            For the CPU device, users need to write a dynamic cluster startup script, please see the `Dynamic Cluster
+            Startup <https://www.mindspore.cn/tutorials/experts/en/master/parallel/dynamic_cluster.html>`_ .
+
         >>> from mindspore.communication import init
         >>> init()
     """
@@ -178,10 +182,6 @@ def init(backend_name=None):
         if device_target != "Ascend":
             raise RuntimeError("For 'init', the argument 'backend_name' should be '{}' to init '{}', "
                                "but got 'hccl'.".format(DEVICE_TO_BACKEND[device_target], device_target))
-        if is_initialized(device_target):
-            logger.warning(f"For 'init' in Ascend backend, the backend is already initialized, please set it before "
-                           "the definition of any Tensor and Parameter, and the instantiation and execution of any "
-                           "operation and net, otherwise the 'init' may not take effect.")
         if not host_init:
             _check_parallel_envs()
         GlobalComm.BACKEND = Backend("hccl")
@@ -224,12 +224,17 @@ def release():
         .. note::
             Before running the following examples, you need to configure the communication environment variables.
 
-            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/tutorials/experts/zh-CN/master/parallel/msrun_launcher.html>`_
+            For the Ascend devices, users need to prepare the rank table, set rank_id and device_id.
+            Please see the `rank table Startup
+            <https://www.mindspore.cn/tutorials/experts/en/master/parallel/rank_table.html>`_
             for more details.
 
+            For the GPU devices, users need to prepare the host file and mpi, please see the `mpirun Startup
+            <https://www.mindspore.cn/tutorials/experts/en/master/parallel/mpirun.html>`_ .
+
+            For the CPU device, users need to write a dynamic cluster startup script, please see the `Dynamic Cluster
+            Startup <https://www.mindspore.cn/tutorials/experts/en/master/parallel/dynamic_cluster.html>`_ .
+
         >>> from mindspore.communication import init, release
         >>> init()
         >>> release()
@@ -263,12 +268,17 @@ def get_rank(group=GlobalComm.WORLD_COMM_GROUP):
         .. note::
             Before running the following examples, you need to configure the communication environment variables.
 
-            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/tutorials/experts/zh-CN/master/parallel/msrun_launcher.html>`_
+            For the Ascend devices, users need to prepare the rank table, set rank_id and device_id.
+            Please see the `rank table Startup
+            <https://www.mindspore.cn/tutorials/experts/en/master/parallel/rank_table.html>`_
             for more details.
 
+            For the GPU devices, users need to prepare the host file and mpi, please see the `mpirun Startup
+            <https://www.mindspore.cn/tutorials/experts/en/master/parallel/mpirun.html>`_ .
+
+            For the CPU device, users need to write a dynamic cluster startup script, please see the `Dynamic Cluster
+            Startup <https://www.mindspore.cn/tutorials/experts/en/master/parallel/dynamic_cluster.html>`_ .
+
         >>> from mindspore.communication import init, get_rank
         >>> init()
         >>> rank_id = get_rank()
@@ -308,12 +318,17 @@ def get_local_rank(group=GlobalComm.WORLD_COMM_GROUP):
         .. note::
             Before running the following examples, you need to configure the communication environment variables.
 
-            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/tutorials/experts/zh-CN/master/parallel/msrun_launcher.html>`_
+            For the Ascend devices, users need to prepare the rank table, set rank_id and device_id.
+            Please see the `rank table Startup
+            <https://www.mindspore.cn/tutorials/experts/en/master/parallel/rank_table.html>`_
             for more details.
 
+            For the GPU devices, users need to prepare the host file and mpi, please see the `mpirun Startup
+            <https://www.mindspore.cn/tutorials/experts/en/master/parallel/mpirun.html>`_ .
+
+            For the CPU device, users need to write a dynamic cluster startup script, please see the `Dynamic Cluster
+            Startup <https://www.mindspore.cn/tutorials/experts/en/master/parallel/dynamic_cluster.html>`_ .
+
         >>> import mindspore as ms
         >>> from mindspore.communication import init, get_rank, get_local_rank
         >>> ms.set_context(device_target="Ascend")
@@ -356,12 +371,17 @@ def get_group_size(group=GlobalComm.WORLD_COMM_GROUP):
         .. note::
             Before running the following examples, you need to configure the communication environment variables.
 
-            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/tutorials/experts/zh-CN/master/parallel/msrun_launcher.html>`_
+            For the Ascend devices, users need to prepare the rank table, set rank_id and device_id.
+            Please see the `rank table Startup
+            <https://www.mindspore.cn/tutorials/experts/en/master/parallel/rank_table.html>`_
             for more details.
 
+            For the GPU devices, users need to prepare the host file and mpi, please see the `mpirun Startup
+            <https://www.mindspore.cn/tutorials/experts/en/master/parallel/mpirun.html>`_ .
+
+            For the CPU device, users need to write a dynamic cluster startup script, please see the `Dynamic Cluster
+            Startup <https://www.mindspore.cn/tutorials/experts/en/master/parallel/dynamic_cluster.html>`_ .
+
         >>> import mindspore as ms
         >>> from mindspore.communication import init, get_group_size
         >>> ms.set_auto_parallel_context(device_num=8)
@@ -403,12 +423,17 @@ def get_local_rank_size(group=GlobalComm.WORLD_COMM_GROUP):
         .. note::
             Before running the following examples, you need to configure the communication environment variables.
 
-            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/tutorials/experts/zh-CN/master/parallel/msrun_launcher.html>`_
+            For the Ascend devices, users need to prepare the rank table, set rank_id and device_id.
+            Please see the `rank table Startup
+            <https://www.mindspore.cn/tutorials/experts/en/master/parallel/rank_table.html>`_
             for more details.
 
+            For the GPU devices, users need to prepare the host file and mpi, please see the `mpirun Startup
+            <https://www.mindspore.cn/tutorials/experts/en/master/parallel/mpirun.html>`_ .
+
+            For the CPU device, users need to write a dynamic cluster startup script, please see the `Dynamic Cluster
+            Startup <https://www.mindspore.cn/tutorials/experts/en/master/parallel/dynamic_cluster.html>`_ .
+
         >>> import mindspore as ms
         >>> from mindspore.communication import init, get_local_rank_size
         >>> ms.set_context(device_target="Ascend")
@@ -453,12 +478,17 @@ def get_world_rank_from_group_rank(group, group_rank_id):
         .. note::
             Before running the following examples, you need to configure the communication environment variables.
 
-            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/tutorials/experts/zh-CN/master/parallel/msrun_launcher.html>`_
+            For the Ascend devices, users need to prepare the rank table, set rank_id and device_id.
+            Please see the `rank table Startup
+            <https://www.mindspore.cn/tutorials/experts/en/master/parallel/rank_table.html>`_
             for more details.
 
+            For the GPU devices, users need to prepare the host file and mpi, please see the `mpirun Startup
+            <https://www.mindspore.cn/tutorials/experts/en/master/parallel/mpirun.html>`_
+
+            For the CPU device, users need to write a dynamic cluster startup script, please see the `Dynamic Cluster
+            Startup <https://www.mindspore.cn/tutorials/experts/en/master/parallel/dynamic_cluster.html>`_ .
+
         >>> import mindspore as ms
         >>> from mindspore import set_context
         >>> from mindspore.communication import init, create_group, get_world_rank_from_group_rank, get_rank
@@ -507,12 +537,17 @@ def get_group_rank_from_world_rank(world_rank_id, group):
         .. note::
             Before running the following examples, you need to configure the communication environment variables.
 
-            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/tutorials/experts/zh-CN/master/parallel/msrun_launcher.html>`_
+            For the Ascend devices, users need to prepare the rank table, set rank_id and device_id.
+            Please see the `rank table Startup
+            <https://www.mindspore.cn/tutorials/experts/en/master/parallel/rank_table.html>`_
             for more details.
 
+            For the GPU devices, users need to prepare the host file and mpi, please see the `mpirun Startup
+            <https://www.mindspore.cn/tutorials/experts/en/master/parallel/mpirun.html>`_
+
+            For the CPU device, users need to write a dynamic cluster startup script, please see the `Dynamic Cluster
+            Startup <https://www.mindspore.cn/tutorials/experts/en/master/parallel/dynamic_cluster.html>`_ .
+
         >>> import mindspore as ms
         >>> from mindspore import set_context
         >>> from mindspore.communication import init, create_group, get_group_rank_from_world_rank, get_rank
@@ -558,15 +593,20 @@ def create_group(group, rank_ids):
         .. note::
             Before running the following examples, you need to configure the communication environment variables.
 
-            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/tutorials/experts/zh-CN/master/parallel/msrun_launcher.html>`_
+            For the Ascend devices, users need to prepare the rank table, set rank_id and device_id.
+            Please see the `rank table Startup
+            <https://www.mindspore.cn/tutorials/experts/en/master/parallel/rank_table.html>`_
             for more details.
 
+            For the GPU devices, users need to prepare the host file and mpi, please see the `mpirun Startup
+            <https://www.mindspore.cn/tutorials/experts/en/master/parallel/mpirun.html>`_ .
+
+            For the CPU device, users need to write a dynamic cluster startup script, please see the `Dynamic Cluster
+            Startup <https://www.mindspore.cn/tutorials/experts/en/master/parallel/dynamic_cluster.html>`_ .
+
         >>> import mindspore as ms
         >>> from mindspore import set_context
-        >>> from mindspore import ops
+        >>> import mindspore.ops as ops
         >>> from mindspore.communication import init, create_group, get_rank
         >>> set_context(mode=ms.GRAPH_MODE, device_target="Ascend")
         >>> init()
@@ -606,15 +646,20 @@ def destroy_group(group):
         .. note::
             Before running the following examples, you need to configure the communication environment variables.
 
-            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/tutorials/experts/zh-CN/master/parallel/msrun_launcher.html>`_
+            For the Ascend devices, users need to prepare the rank table, set rank_id and device_id.
+            Please see the `rank table startup
+            <https://www.mindspore.cn/tutorials/experts/en/master/parallel/rank_table.html>`_
             for more details.
 
+            For the GPU devices, users need to prepare the host file and mpi, please see the `mpirun startup
+            <https://www.mindspore.cn/tutorials/experts/en/master/parallel/mpirun.html>`_ .
+
+            For the CPU device, users need to write a dynamic cluster startup script, please see the `Dynamic Cluster
+            Startup <https://www.mindspore.cn/tutorials/experts/en/master/parallel/dynamic_cluster.html>`_ .
+
         >>> import mindspore as ms
         >>> from mindspore import set_context
-        >>> from mindspore import ops
+        >>> import mindspore.ops as ops
         >>> from mindspore.communication import init, create_group, destroy_group, get_rank
         >>> set_context(mode=ms.GRAPH_MODE, device_target="Ascend")
         >>> init()
@@ -628,46 +673,3 @@ def destroy_group(group):
         raise TypeError("For 'destroy_group', the argument 'group' must be type of string, "
                         "but got 'group' type : {}.".format(type(group)))
     _destroy_group_helper(group)
-
-
-def get_process_group_ranks(group=GlobalComm.WORLD_COMM_GROUP):
-    """
-    Gets the ranks of the specific group and returns the process ranks in the communication group as a list.
-
-    Args:
-        group (str, optional): The communication group to work on. Default: ``GlobalComm.WORLD_COMM_GROUP`` , which
-            means ``"hccl_world_group"`` in Ascend, and ``"nccl_world_group"`` in GPU.
-
-    Returns:
-        List (List[int]), List of process ranks in the specified communication group.
-
-    Raises:
-        TypeError: If the `group` is not a str.
-        RuntimeError: If device target is invalid, or backend is invalid, or distributed initialization fails.
-
-    Supported Platforms:
-        ``Ascend`` ``GPU``
-
-    Examples:
-        .. note::
-            Before running the following examples, you need to configure the communication environment variables.
-
-            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/tutorials/experts/zh-CN/master/parallel/msrun_launcher.html>`_
-            for more details.
-
-            This example should be run with 4 devices.
-
-        >>> import numpy as np
-        >>> from mindspore.communication import init, get_process_group_ranks
-        >>>
-        >>> init()
-        >>> output = get_process_group_ranks()
-        >>> print(output)
-        [0, 1, 2, 3]
-
-    """
-    return _get_group_ranks(group)