mindspore 2.4.10__cp39-cp39-win_amd64.whl → 2.6.0rc1__cp39-cp39-win_amd64.whl

mindspore/train/summary/_summary_adapter.py

@@ -25,7 +25,7 @@ from PIL import Image
 
 from mindspore import log as logger
 from mindspore import context
-from mindspore.communication.management import get_rank
+from mindspore.communication.management import get_rank, get_group_size
 from mindspore.communication.management import GlobalComm
 
 from mindspore import _checkparam as Validator
@@ -60,7 +60,7 @@ def get_event_file_name(prefix, suffix, time_second):
     file_name = ""
     hostname = platform.node()
 
-    device_num = context.get_auto_parallel_context('device_num')
+    device_num = get_group_size()
     device_id = context.get_context('device_id')
     if device_num > 1 or GlobalComm.WORLD_COMM_GROUP == 'nccl_world_group':
         # Notice:

mindspore/train/summary/summary_record.py

@@ -26,7 +26,7 @@ from mindspore import log as logger
 from mindspore.nn import Cell
 from mindspore import context
 from mindspore._c_expression import security
-from mindspore._c_expression import Tensor as Tensor_
+from mindspore._c_expression import TensorPy as Tensor_
 from mindspore.common import dtype as mstype
 from mindspore.common.tensor import Tensor
 from mindspore import _checkparam as Validator
@@ -393,8 +393,7 @@ class SummaryRecord:
             bool, whether the record process is successful or not.
 
         Raises:
-            TypeError: `step` is not int, or `train_network` is not `mindspore.nn.Cell
-                <https://www.mindspore.cn/docs/en/master/api_python/nn/mindspore.nn.Cell.html#mindspore-nn-cell>`_ .
+            TypeError: `step` is not int, or `train_network` is not :class:`mindspore.nn.Cell`.
 
         Examples:
             >>> import mindspore as ms

mindspore/train/train_thor/model_thor.py

@@ -96,7 +96,7 @@ class ModelThor(Model):
                              elements, including the positions of loss value, predicted value and label. The loss
                              value would be passed to the `Loss` metric, the predicted value and label would be passed
                              to other metric. Default: ``None``.
-        amp_level (str): Option for argument `level` in `mindspore.amp.build_train_network`, level for mixed
+        amp_level (str): Option for argument `level` in :func:`mindspore.amp.build_train_network`, level for mixed
             precision training. Supports [O0, O2, O3]. Default: "O0".
 
             - O0: Do not change.

mindspore/utils/__init__.py

@@ -14,8 +14,11 @@
 # ============================================================================
 """Utils module."""
 from __future__ import absolute_import
-from mindspore._c_expression import stress_detect
-from .utils import ExitByRequest
+from mindspore._c_expression import stress_detect, _reuse_data_ptr
+from .utils import ExitByRequest, _tft_handler
+from .runtime_execution_order_check import runtime_execution_order_check
+from . import dryrun
 
 # Symbols from utils module.
-__all__ = ["stress_detect", "ExitByRequest"]
+__all__ = ["stress_detect", "ExitByRequest", "runtime_execution_order_check", "dryrun", "_reuse_data_ptr",
+           "_tft_handler"]

mindspore/utils/dryrun.py

@@ -0,0 +1,140 @@
+# Copyright 2024 Huawei Technologies Co., Ltd
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+# http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+# ============================================================================
+"""dryrun."""
+import traceback
+import os
+from mindspore.common import Tensor
+from mindspore import log as logger
+from mindspore.common._stub_tensor import StubTensor
+from mindspore.common import dtype as mstype
+from mindspore._checkparam import is_stub_tensor
+
+
+class TraceBack():
+    """
+    traceback warning logs in dryrun mode
+    """
+    def __init__(self):
+        self.stack_str_set = set()
+
+    def inject(self, method):
+        """
+        inject warning logs in dryrun mode
+        """
+        def new_method(*args, **kwargs):
+            stack_list = traceback.format_list(traceback.extract_stack())
+            stack_str = "".join(stack_list)
+            if "Parameter" not in stack_str and stack_str not in self.stack_str_set:
+                self.stack_str_set.add(stack_str)
+                logger.warning("In dryrun mode, you cannot obtain real tensor value, and the traceback is {%s}",
+                               stack_list)
+            return method(*args, **kwargs)
+        return new_method
+
+
+def no_inject_traceback_for_print(self):
+    if is_stub_tensor(self):
+        self = self.stub_sync()
+    if self.dtype == mstype.type_none:
+        return "Unknown Tensor type!"
+    if self.has_init:
+        self.init_data()
+    return str(Tensor.asnumpy(self))
+
+
+def set_simulation():
+    """
+    This interface is used to enable the dryrun function. The dryrun function is mainly used to simulate the actual
+    operation of the large model. After it is enabled, the memory usage, compilation information, etc. can be simulated
+    without occupying device card. In the PyNative mode, once it is enabled, if values are fetched from the device to
+    the host, the Python call stack log will be printed to inform users that these values are inaccurate.
+
+    Supported Platforms:
+        ``Ascend``
+
+    Examples:
+        >>> import mindspore as ms
+        >>> from mindspore.utils import dryrun
+        >>> import numpy as np
+        >>> dryrun.set_simulation()
+        >>> print(os.environ.get('MS_SIMULATION_LEVEL'))
+        1
+    """
+    os.environ["MS_SIMULATION_LEVEL"] = "1"
+    obj = TraceBack()
+    Tensor.asnumpy = obj.inject(Tensor.asnumpy)
+    Tensor._getitem = obj.inject(Tensor._getitem)
+    Tensor.is_contiguous = obj.inject(Tensor.is_contiguous)
+    Tensor.flush_from_cache = obj.inject(Tensor.flush_from_cache)
+    StubTensor.asnumpy = obj.inject(StubTensor.asnumpy)
+    StubTensor._getitem = obj.inject(StubTensor._getitem)
+    StubTensor.is_contiguous = obj.inject(StubTensor.is_contiguous)
+    StubTensor.flush_from_cache = obj.inject(StubTensor.flush_from_cache)
+    Tensor.__str__ = no_inject_traceback_for_print
+    StubTensor.__str__ = no_inject_traceback_for_print
+    Tensor.tolist = obj.inject(Tensor.tolist)
+    Tensor.__int__ = obj.inject(Tensor.__int__)
+    Tensor.__float__ = obj.inject(Tensor.__float__)
+
+
+def mock(mock_val, *args):
+    """
+    In the network, if some `if` branch need to use the actual execution values and the virtual execution cannot obtain
+    them, this interface can be used to return simulated values. During actual execution, the correct results can be
+    obtained and the execution values can be returned.
+
+    Args:
+        mock_val (Union[Value, Tensor]): The value you want to return.
+        args (Union[Value, function]): The content you want to mock, it can be values, function and so on.
+
+    Returns:
+        If dryrun is enabled, mock_val will be returned; otherwise, the actual execution value of args will be returned.
+
+    Supported Platforms:
+        ``Ascend`` ``GPU`` ``CPU``
+
+    Examples:
+        >>> import mindspore as ms
+        >>> from mindspore.utils import dryrun
+        >>> import numpy as np
+        >>> dryrun.set_simulation()
+        >>> a = ms.Tensor(np.random.rand(3, 3).astype(np.float32))
+        >>> if dryrun.mock(True, a[0, 0] > 0.5):
+        ...     print("return mock_val: True.")
+        return mock_val: True
+        >>>
+        >>> import mindspore as ms
+        >>> from mindspore.utils import dryrun
+        >>> import numpy as np
+        >>> a = ms.Tensor(np.ones((3, 3)).astype(np.float32))
+        >>> if dryrun.mock(False, a[0, 0] > 0.5):
+        ...     print("return real execution: True.")
+        return real execution: True.
+        >>>
+        >>> import mindspore as ms
+        >>> from mindspore.utils import dryrun
+        >>> import numpy as np
+        >>> a = ms.Tensor(np.ones((3, 3)).astype(np.float32))
+        >>> if dryrun.mock(False, (a > 0.5).any):
+        ...     print("return real execution: True.")
+        return real execution: True.
+    """
+    if os.environ.get('MS_SIMULATION_LEVEL'):
+        return mock_val
+    if len(args) == 1:
+        if callable(args[0]):
+            return args[0]()
+        return args[0]
+    return args[0](*args[1:])

mindspore/utils/hooks.py

@@ -0,0 +1,81 @@
+# Copyright 2025 Huawei Technologies Co., Ltd
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+# http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+# ============================================================================
+"""hooks"""
+from collections import OrderedDict
+import weakref
+from typing import Any, Tuple
+
+
+class _RemovableHandle:
+    r"""
+    A handle which provides the capability to remove a hook.
+
+    Args:
+        hooks_dict (dict): A dictionary of hooks, indexed by hook `id`.
+
+    Keyword Args:
+        extra_dict (Union[dict, list[dict]], optional): An additional dictionary or list of
+            dictionaries whose keys will be deleted when the same keys are
+            removed from `hooks_dict`. Default ``None``.
+    """
+
+    id: int
+    next_id: int = 0
+
+    def __init__(self, hooks_dict: Any, *, extra_dict: Any = None) -> None:
+        self.hooks_dict_ref = weakref.ref(hooks_dict)
+        self.id = _RemovableHandle.next_id
+        _RemovableHandle.next_id += 1
+
+        self.extra_dict_ref: Tuple = ()
+        if isinstance(extra_dict, dict):
+            self.extra_dict_ref = (weakref.ref(extra_dict),)
+        elif isinstance(extra_dict, list):
+            self.extra_dict_ref = tuple(weakref.ref(d) for d in extra_dict)
+
+    def remove(self) -> None:
+        hooks_dict = self.hooks_dict_ref()
+        if hooks_dict is not None and self.id in hooks_dict:
+            del hooks_dict[self.id]
+
+        for ref in self.extra_dict_ref:
+            extra_dict = ref()
+            if extra_dict is not None and self.id in extra_dict:
+                del extra_dict[self.id]
+
+    def __getstate__(self):
+        if self.extra_dict_ref is None:
+            return (self.hooks_dict_ref(), self.id)
+        return (self.hooks_dict_ref(), self.id, tuple(ref() for ref in self.extra_dict_ref))
+
+    def __setstate__(self, state) -> None:
+        if state[0] is None:
+            # create a dead reference
+            self.hooks_dict_ref = weakref.ref(OrderedDict())
+        else:
+            self.hooks_dict_ref = weakref.ref(state[0])
+        self.id = state[1]
+        _RemovableHandle.next_id = max(_RemovableHandle.next_id, self.id + 1)
+
+        if len(state) < 3 or state[2] is None:
+            self.extra_dict_ref = ()
+        else:
+            self.extra_dict_ref = tuple(weakref.ref(d) for d in state[2])
+
+    def __enter__(self) -> "_RemovableHandle":
+        return self
+
+    def __exit__(self, type: Any, value: Any, tb: Any) -> None:
+        self.remove()