mindspore 2.7.0rc1__cp310-cp310-win_amd64.whl → 2.7.1__cp310-cp310-win_amd64.whl

mindspore/onnx/onnx_export.py

@@ -0,0 +1,137 @@
+# 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.
+# ============================================================================
+
+"""Model export to ONNX."""
+from __future__ import absolute_import
+from __future__ import division
+
+import os
+
+import mindspore.nn as nn
+from mindspore import log as logger
+from mindspore._checkparam import check_input_dataset
+from mindspore import _checkparam as Validator
+from mindspore.common.api import _cell_graph_executor as _executor
+from mindspore.train.serialization import _calculation_net_size
+from mindspore.dataset.engine.datasets import Dataset
+
+PROTO_LIMIT_SIZE = 1024 * 1024 * 2
+
+
+def export(net, *inputs, file_name, input_names=None, output_names=None, export_params=True,
+           keep_initializers_as_inputs=False, dynamic_axes=None):
+    """
+    Export the MindSpore network into an ONNX model.
+
+    Note:
+        - Support exporting network larger than 2GB. When the network exceeds 2GB,
+          parameters are saved in additional binary files stored in the same directory as the ONNX file.
+        - When `file_name` does not have a suffix, the system will automatically add the suffix `.onnx` .
+
+    Args:
+        net (Union[Cell, function]): MindSpore network.
+        inputs (Union[Tensor, list, tuple, Number, bool]): It represents the inputs of the `net` , if the network has
+            multiple inputs, set them together.
+        file_name (str): File name of the model to be exported.
+        input_names (list, optional): Names to assign to the input nodes of the graph, in order. Default: ``None`` .
+        output_names (list, optional): Names to assign to the output nodes of the graph, in order. Default: ``None`` .
+        export_params (bool, optional): If false, parameters (weights) will not be exported,
+            parameters will add input nodes as input of the graph. Default: ``True`` .
+        keep_initializers_as_inputs (bool, optional): If True, all the initializers (model parameters/weights) will
+            add as inputs to the graph. This allows modifying any or all weights when running the exported ONNX model.
+            Default: ``False`` .
+        dynamic_axes (dict[str, dict[int, str]], optional): To specify axes of input tensors as dynamic (at runtime).
+            Default: ``None`` .
+
+            - Set a dict with scheme: {input_node_name: {axis_index:axis_name}},
+              for example, {"input1": {0:"batch_size", 1: "seq_len"}, "input2": {{0:"batch_size"}}.
+            - By default, the shapes of all input tensors in the exported model exactly match those specified in
+              `inputs`.
+
+    Raises:
+        ValueError: If the parameter `net` is not :class:`mindspore.nn.Cell`.
+        ValueError: If the parameter `input_names` is not list type.
+        ValueError: If the parameter `output_names` is not list type
+        ValueError: If the parameter `dynamic_axes` is not dict type.
+
+    Examples:
+        >>> import mindspore as ms
+        >>> import numpy as np
+        >>> from mindspore import Tensor
+        >>>
+        >>> # Define the network structure of LeNet5. Refer to
+        >>> # https://gitee.com/mindspore/docs/blob/master/docs/mindspore/code/lenet.py
+        >>> net = LeNet5()
+        >>> input_tensor = Tensor(np.ones([1, 1, 32, 32]).astype(np.float32))
+        >>> ms.onnx.export(net, input_tensor, file_name='lenet.onnx', input_names=['input1'], output_names=['output1'])
+
+    """
+    Validator.check_file_name_by_regular(file_name)
+    logger.info("exporting model file:%s format:%s.", file_name, "ONNX")
+    Validator.check_isinstance("net", net, nn.Cell)
+    input_names = input_names or []
+    Validator.check_isinstance("input_names", input_names, list)
+    output_names = output_names or []
+    Validator.check_isinstance("output_names", output_names, list)
+    dynamic_axes = dynamic_axes or {}
+    Validator.check_isinstance("dynamic_axes", dynamic_axes, dict)
+
+    if check_input_dataset(*inputs, dataset_type=Dataset):
+        raise ValueError(f"Can not support dataset as inputs to export ONNX model.")
+
+    cell_mode = net.training
+    net.set_train(mode=False)
+
+    extra_save_params = False
+    total_size = _calculation_net_size(net)
+    if total_size > PROTO_LIMIT_SIZE:
+        logger.warning('Network size is: {}G, it exceeded the protobuf: {}G limit, now parameters in network are saved '
+                       'in external data files.'.format(total_size / 1024 / 1024, PROTO_LIMIT_SIZE / 1024 / 1024))
+        extra_save_params = True
+
+    phase_name = 'export.onnx'
+    graph_id, _ = _executor.compile(net, *inputs, phase=phase_name, do_convert=False)
+
+    abs_file_name = os.path.abspath(file_name)
+    if not abs_file_name.endswith('.onnx'):
+        abs_file_name += ".onnx"
+
+    dir_path = os.path.dirname(abs_file_name)
+    if dir_path and not os.path.exists(dir_path):
+        os.makedirs(dir_path, exist_ok=True)
+
+    abs_file_dir = os.path.dirname(abs_file_name) if extra_save_params else ""
+
+    onnx_stream = _executor._get_onnx_func_graph_proto(obj=net, exec_id=graph_id, input_names=input_names,
+                                                       output_names=output_names, export_params=export_params,
+                                                       keep_initializers_as_inputs=keep_initializers_as_inputs,
+                                                       dynamic_axes=dynamic_axes, extra_save_params=extra_save_params,
+                                                       save_file_dir=abs_file_dir)
+    if onnx_stream is None:
+        raise RuntimeError("Export onnx model failed, ensure that the model has been compiled correctly")
+
+    try:
+        with open(abs_file_name, 'wb') as f:
+            f.write(onnx_stream)
+
+        if os.path.getsize(abs_file_name) != len(onnx_stream):
+            logger.warning("ONNX file size doesn't match expected value, but proceeding continue.")
+
+    except IOError as e:
+        logger.error(f"Failed to write ONNX file: {e}")
+        if os.path.exists(abs_file_name):
+            os.remove(abs_file_name)
+
+    net.set_train(mode=cell_mode)

mindspore/ops/__init__.py

@@ -37,6 +37,7 @@ from mindspore.ops.functional_overload import all_gather_matmul, matmul_reduce_s
 from mindspore.ops.composite import *
 from mindspore.ops.operations import *
 from mindspore.ops.function import *
+from mindspore.ops.communication import *
 from mindspore.ops.functional import *
 from mindspore.ops._utils import arg_dtype_cast, arg_handler
 
@@ -55,4 +56,5 @@ __all__.extend(composite.__all__)
 __all__.extend(operations.__all__)
 __all__.extend(functional.__all__)
 __all__.extend(function.__all__)
+__all__.extend(communication.__all__)
 __all__.extend(auto_generate.__all__)

mindspore/ops/_grad_experimental/grad_comm_ops.py

@@ -29,7 +29,7 @@ from mindspore.common.sparse_tensor import RowTensorInner
 from mindspore.ops.composite.multitype_ops.zeros_like_impl import zeros_like
 from mindspore.ops.operations.comm_ops import (AllGather, _MiniStepAllGather, _HostAllGather, AllReduce,
                                                NeighborExchange, AlltoAll, AlltoAllV, NeighborExchangeV2,
-                                               Broadcast, AllGatherV, ReduceScatterV,
+                                               Broadcast, AlltoAllVC, AllGatherV, ReduceScatterV,
                                                _GetTensorSlice, _MirrorOperator, _MirrorMiniStepOperator, ReduceOp,
                                                ReduceScatter, _HostReduceScatter, _VirtualDiv, _VirtualAdd, _AllSwap,
                                                _VirtualAssignAdd, _VirtualAccuGrad, _MirrorMicroStepOperator,
@@ -37,6 +37,7 @@ from mindspore.ops.operations.comm_ops import (AllGather, _MiniStepAllGather, _H
                                                _VirtualAssignKvCache)
 from mindspore.ops._grad_experimental.grad_base import bprop_getters
 from mindspore.ops.operations import _grad_ops as G
+from mindspore.tools.sdc_detect import _sdc_detector
 import mindspore as ms
 
 _squared_device_local_norm = None
@@ -277,6 +278,8 @@ def get_bprop_mirror_micro_step_operator(self):
     if dump_device_local_norm:
         # init _squared _squared_device_local_norm
         squared_device_local_norm = get_squared_device_local_norm_param()
+    # feature value sampling for sdc detect
+    feat_value_dump_name = _sdc_detector.get_dump_name(param_name) if _sdc_detector.need_sample() else None
 
     def bprop(x, z, out, dout):
         if dump_local_norm or dump_device_local_norm:
@@ -289,6 +292,9 @@ def get_bprop_mirror_micro_step_operator(self):
             if dump_device_local_norm:
                 z = F.depend(z, F.assign_add(squared_device_local_norm,
                                              cast(squared_norm, squared_device_local_norm.dtype)))
+        if feat_value_dump_name and z.ndim > 1:
+            feat_value = square(F.max(F.abs(z))[0])
+            z = F.depend(z, tensor_dump(feat_value_dump_name, feat_value))
         real_grad = z
         assign_out = dout
         if issubclass_(F.typeof(dout), mstype.tensor_type):
@@ -343,14 +349,16 @@ def get_bprop_all_gather(self):
         ln_print = P.Print()
         tensor_dump = P.TensorDump()
         reduce_sum = P.ReduceSum(keep_dims=False)
-        square = P.Square()
         sqrt = P.Sqrt()
+    square = P.Square()
     if dump_local_norm_path:
         global_rank = get_rank()
         file = os.path.join(dump_local_norm_path, "rank_" + str(global_rank), "local_norm__" + param_name)
     if dump_device_local_norm:
         # init _squared _squared_device_local_norm
         squared_device_local_norm = get_squared_device_local_norm_param()
+    # feature value sampling for sdc detect
+    feat_value_dump_name = _sdc_detector.get_dump_name(param_name) if _sdc_detector.need_sample() else None
 
     def bprop(x, out, dout):
         if param_name and (dump_local_norm or dump_device_local_norm):
@@ -363,6 +371,9 @@ def get_bprop_all_gather(self):
             if dump_device_local_norm:
                 dout = F.depend(dout, F.assign_add(squared_device_local_norm,
                                                    cast(squared_norm, squared_device_local_norm.dtype)))
+        if param_name and feat_value_dump_name and dout.ndim > 1:
+            feat_value = square(F.max(F.abs(dout))[0])
+            dout = F.depend(dout, tensor_dump(feat_value_dump_name, feat_value))
 
         dx = reduce_scatter(dout)
         if mean_flag:
@@ -452,6 +463,8 @@ def get_bprop_micro_step_all_gather(self):
     if dump_device_local_norm:
         # init _squared _squared_device_local_norm
         squared_device_local_norm = get_squared_device_local_norm_param()
+    # feature value sampling for sdc detect
+    feat_value_dump_name = _sdc_detector.get_dump_name(param_name) if _sdc_detector.need_sample() else None
 
     def bprop(x, z, out, dout):
         if with_mirror_operator:
@@ -472,6 +485,9 @@ def get_bprop_micro_step_all_gather(self):
             if dump_device_local_norm:
                 z = F.depend(z, F.assign_add(squared_device_local_norm,
                                              cast(squared_norm, squared_device_local_norm.dtype)))
+        if feat_value_dump_name and z.ndim > 1:
+            feat_value = square(F.max(F.abs(z))[0])
+            z = F.depend(z, tensor_dump(feat_value_dump_name, feat_value))
         if not do_mirror:
             return (z, cast(out_tensor, dtype(z)))
         real_grad = reduce_scatter(z)
@@ -655,6 +671,21 @@ def get_bprop_all_to_all_v(self):
     return bprop
 
 
+@bprop_getters.register(AlltoAllVC)
+def get_bprop_all_to_all_v_c(self):
+    """Generate bprop for AlltoAllVC."""
+    all_to_all_v_c_grad = AlltoAllVC(self.group, self.block_size, transpose=True)
+    if hasattr(self, "instance_name") and self.instance_name:
+        instance_name = "grad" + self.instance_name
+        all_to_all_v_c_grad.set_prim_instance_name(instance_name)
+
+    def bprop(x, send_count_matrix, out, dout):
+        dx = all_to_all_v_c_grad(dout, send_count_matrix)
+        return (dx, zeros_like(send_count_matrix))
+
+    return bprop
+
+
 @bprop_getters.register(AllGatherV)
 def get_bprop_all_gather_v(self):
     """Generate bprop for AllGatherV."""
@@ -728,6 +759,8 @@ def get_bprop_mirror_operator(self):
     if dump_device_local_norm:
         # init _squared _squared_device_local_norm
         squared_device_local_norm = get_squared_device_local_norm_param()
+    # feature value sampling for sdc detect
+    feat_value_dump_name = _sdc_detector.get_dump_name(param_name) if _sdc_detector.need_sample() else None
     if dev_num > 1:
         dev_num_r = 1.0 / dev_num
         all_reduce = AllReduce(group=group)
@@ -762,6 +795,9 @@ def get_bprop_mirror_operator(self):
             if dump_device_local_norm:
                 dout = F.depend(dout, F.assign_add(squared_device_local_norm,
                                                    cast(squared_norm, squared_device_local_norm.dtype)))
+        if feat_value_dump_name and dout.ndim > 1:
+            feat_value = square(F.max(F.abs(dout))[0])
+            dout = F.depend(dout, tensor_dump(feat_value_dump_name, feat_value))
 
         if dev_num == 1:
             return (dout,)

mindspore/ops/_grad_experimental/grad_inner_ops.py

@@ -23,15 +23,6 @@ from mindspore.ops.composite.multitype_ops.zeros_like_impl import zeros_like
 from mindspore.ops._grad_experimental.grad_base import bprop_getters
 
 
-@bprop_getters.register("raise")
-def get_bprop_raise(self):
-    """Grad definition for `raise` operation."""
-    def bprop(x, y, z, out, dout):
-        return x, y, z
-
-    return bprop
-
-
 @bprop_getters.register(inner.ParallelResizeBilinear)
 def get_bprop_parallel_resize_bilinear(self):
     """Grad definition for `ParallelResizeBilinear` operation."""

mindspore/ops/_op_impl/aicpu/__init__.py

@@ -214,13 +214,8 @@ from .cumsum import _cumsum_aicpu
 from .round import _round_aicpu
 from .stft import _stft_aicpu
 from .floor_div import _floor_div_aicpu
-from .priority_replay_buffer import _prb_create_op_cpu
-from .priority_replay_buffer import _prb_push_op_cpu
 from .conjugate_transpose import _conjugate_transpose_aicpu
-from .priority_replay_buffer import _prb_sample_op_cpu
-from .priority_replay_buffer import _prb_update_op_cpu
 from .equal import _equal_aicpu
-from .priority_replay_buffer import _prb_destroy_op_cpu
 from .right_shift import _right_shift_aicpu
 from .tril import _tril_aicpu
 from .linspace import _lin_space_aicpu
@@ -242,10 +237,6 @@ from .sparse_tensor_to_csr_sparse_matrix import _sparse_tensor_to_csr_sparse_mat
 from .csr_sparse_matrix_to_sparse_tensor import _csr_sparse_matrix_to_sparse_tensor_aicpu
 from .linear_sum_assignment import _linear_sum_assignment_aicpu
 from .random_shuffle import _random_shuffle_aicpu
-from .reservoir_replay_buffer import _rrb_create_op_cpu
-from .reservoir_replay_buffer import _rrb_push_op_cpu
-from .reservoir_replay_buffer import _rrb_sample_op_cpu
-from .reservoir_replay_buffer import _rrb_destroy_op_cpu
 from .concat_offset import _concat_offset_aicpu
 from .range import _range_aicpu
 from .range_v2 import _range_v2_aicpu
@@ -414,7 +405,6 @@ from .segment_prod import _segment_prod_aicpu
 from .segment_sum import _segment_sum_aicpu
 from .set_size import _set_size_aicpu
 from .slice import _slice_aicpu
-from .sparse_cross import _sparse_cross_aicpu
 from .sparse_slice import _sparse_slice_aicpu
 from .sparse_softmax import _sparse_softmax_aicpu
 from .sparse_tensor_dense_add import _sparse_tensor_dense_add_aicpu

mindspore/ops/_op_impl/cpu/__init__.py

@@ -69,11 +69,7 @@ from .tensor_copy_slices import _tensor_copy_slices_cpu
 from .l2loss import _l2loss_cpu
 from .pyexecute import _pyexecute_cpu
 from .pyfunc import _pyfunc_cpu
-from .buffer_append import _buffer_append_cpu
-from .buffer_get import _buffer_get_cpu
 from .raise_op import _raise_cpu
-from .buffer_sample import _buffer_sample_cpu
-from .priority_replay_buffer import _prb_push_op_cpu
-from .priority_replay_buffer import _prb_sample_op_cpu
+from .joinedstr_op import _joinedstr_cpu
 from .space_to_batch_nd import _space_to_batch_nd_cpu
 from .sspaddmm import _sspaddmm_cpu

mindspore/ops/_op_impl/cpu/{buffer_append.py → joinedstr_op.py}

@@ -1,4 +1,4 @@
-# Copyright 2021 Huawei Technologies Co., Ltd
+# 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.
@@ -11,18 +11,18 @@
 # 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.
-
-"""BufferAppend op"""
+# ============================================================================
+"""JoinedStr op"""
 from mindspore.ops.op_info_register import op_info_register, CpuRegOp, DataType
 
-buffer_append_op_info = CpuRegOp("BufferAppend") \
+joinedstr_op_info = CpuRegOp("JoinedStr") \
     .input(0, "x", "dynamic") \
-    .output(0, "output", "dynamic") \
+    .output(0, "y", "dynamic") \
     .dtype_format(DataType.I32_Default, DataType.I32_Default) \
     .get_op_info()
 
 
-@op_info_register(buffer_append_op_info)
-def _buffer_append_cpu():
-    """BufferAppend cpu register"""
+@op_info_register(joinedstr_op_info)
+def _joinedstr_cpu():
+    """JoinedStr cpu register"""
     return

mindspore/ops/auto_generate/cpp_create_prim_instance_helper.py

@@ -96,6 +96,8 @@ op_args_default_value = {
     "ConvTranspose2D": {"bias": None, "stride": 1, "padding": 0, "output_padding": 0, "groups": 1, "dilation": 1},
     "Correlate": {"pad_mode": 'valid'},
     "CountNonZero": {"dim": None},
+    "CrossEntropyLossGrad": {"weight": None, "grad_zloss": None, "lse_for_zloss": None, "reduction": 'mean', "ignore_index": -100, "label_smoothing": 0.0, "lse_square_scale_for_zloss": 0.0},
+    "CrossEntropyLoss": {"weight": None, "reduction": 'mean', "ignore_index": -100, "label_smoothing": 0.0, "lse_square_scale_for_zloss": 0.0, "return_zloss": False},
     "Cross": {"dim": -65530},
     "CumsumExt": {"dtype": None},
     "CumProd": {"exclusive": False, "reverse": False},
@@ -103,6 +105,7 @@ op_args_default_value = {
     "DCTN": {"type": 2, "s": None, "axes": None, "norm": None},
     "DCT": {"type": 2, "n": None, "axis": -1, "norm": None},
     "Dense": {"bias": None},
+    "DequantSwigluQuant": {"bias": None, "quant_scale": None, "quant_offset": None, "group_index": None, "activate_left": False, "quant_mode": 'static'},
     "Diagonal": {"offset": 0, "dim1": 0, "dim2": 1},
     "DiagonalView": {"offset": 0, "dim1": 0, "dim2": 1},
     "DiagExt": {"diagonal": 0},
@@ -113,20 +116,10 @@ op_args_default_value = {
     "EluExt": {"alpha": 1.0},
     "EluGradExt": {"alpha": 1.0, "is_result": False},
     "Elu": {"alpha": 1.0},
-    "EmbeddingApplyAdamW": {"ams_grad": (0,), "mask_zero": (0,), "padding_key": (0,), "padding_key_mask": (1,), "completion_key": (0,), "completion_key_mask": (1,), "_embedding_dim": 1, "_max_key_num": 1},
-    "EmbeddingApplyAdam": {"mask_zero": (0,), "padding_key": (0,), "padding_key_mask": (1,), "completion_key": (0,), "completion_key_mask": (1,), "_embedding_dim": 1, "_max_key_num": 1},
-    "EmbeddingApplyAdaGrad": {"mask_zero": (0,), "padding_key": (0,), "padding_key_mask": (1,), "completion_key": (0,), "completion_key_mask": (1,), "_embedding_dim": 1, "_max_key_num": 1},
-    "EmbeddingApplyFtrl": {"mask_zero": (0,), "padding_key": (0,), "padding_key_mask": (1,), "completion_key": (0,), "completion_key_mask": (1,), "_embedding_dim": 1, "_max_key_num": 1},
-    "EmbeddingApplyRmsprop": {"mask_zero": (0,), "padding_key": (0,), "padding_key_mask": (1,), "completion_key": (0,), "completion_key_mask": (1,), "_embedding_dim": 1, "_max_key_num": 1},
-    "EmbeddingApplySgd": {"mask_zero": (0,), "padding_key": (0,), "padding_key_mask": (1,), "completion_key": (0,), "completion_key_mask": (1,), "_embedding_dim": 1, "_max_key_num": 1},
     "EmbeddingDenseBackward": {"padding_idx": None, "scale_grad_by_freq": False},
-    "EmbeddingFeatureMappingFileSize": {"only_offset_flag": True},
-    "EmbeddingFeatureMappingFind": {"num": 1},
-    "EmbeddingFeatureMappingImport": {"only_offset_flag": True, "num": 1},
     "Embedding": {"padding_idx": None, "max_norm": None, "norm_type": 2.0, "scale_grad_by_freq": False},
-    "EmbeddingTableEvict": {"steps_to_live": 0},
-    "EmptyLike": {"dtype": None, "device": None},
-    "Empty": {"dtype": None, "device": None},
+    "EmptyLike": {"dtype": None, "device": None, "pin_memory": False},
+    "Empty": {"dtype": None, "device": None, "pin_memory": False},
     "ExtractImagePatches": {"padding": 'VALID'},
     "FFNExt": {"expertTokens": None, "bias1": None, "bias2": None, "scale": None, "offset": None, "deqScale1": None, "deqScale2": None, "antiquant_scale1": None, "antiquant_scale2": None, "antiquant_offset1": None, "antiquant_offset2": None, "activation": 'fastgelu', "inner_precise": 0},
     "FFT2": {"s": None, "dim": (-2, -1), "norm": None},
@@ -177,11 +170,13 @@ op_args_default_value = {
     "IndexAddExt": {"alpha": 1},
     "InnerInplaceIndexPut": {"accumulate": False},
     "InnerMoeTokenUnpermute": {"probs": None, "padded_mode": False, "restore_shape": None},
+    "InnerUnique": {"sorted": True, "return_inverse": False},
     "InplaceAddmm": {"beta": 1, "alpha": 1},
     "InplaceAddsExt": {"alpha": 1},
     "InplaceAddExt": {"alpha": 1},
     "InplaceClampScalar": {"min": None, "max": None},
     "InplaceClampTensor": {"min": None, "max": None},
+    "InplaceCopy": {"non_blocking": False},
     "InplaceDivMods": {"rounding_mode": None},
     "InplaceDivMod": {"rounding_mode": None},
     "InplaceElu": {"alpha": 1.0},
@@ -239,6 +234,8 @@ op_args_default_value = {
     "Meshgrid": {"indexing": 'xy'},
     "MinimumGrad": {"grad_x": True, "grad_y": True},
     "MinDim": {"keepdim": False},
+    "Mla": {"attn_mask": None, "deq_scale_qk": None, "deq_scale_pv": None, "q_seq_lens": None, "context_lens": None, "head_num": 32, "scale_value": 0.0, "kv_head_num": 1, "mask_mode": 'MASK_NONE', "is_ring": 0},
+    "MlaPreprocess": {"param_cache_mode": 0},
     "MoeDistributeCombine": {"tp_send_counts": None, "x_active_mask": None, "activate_scale": None, "weight_scale": None, "group_list": None, "expand_scales": None, "group_ep": None, "group_tp": None, "tp_world_size": 0, "tp_rank_id": 0, "expert_shard_type": 0, "shared_expert_num": 0, "shared_export_rank_num": 0, "global_bs": 0, "out_dtype": 0, "common_quant_mode": 0, "group_list_type": 0},
     "MoeDistributeDispatch": {"expert_scales": None, "scales": None, "x_active_mask": None, "group_ep": None, "group_tp": None, "tp_world_size": 0, "tp_rank_id": 0, "expert_shard_type": 0, "shared_expert_num": 0, "shared_expert_rank_num": 0, "quant_mode": 0, "global_bs": 0, "expert_token_nums_type": 0},
     "MoeTokenPermuteGrad": {"num_topk": 1, "padded_mode": False},
@@ -300,6 +297,7 @@ op_args_default_value = {
     "RFFTFreq": {"d": 1.0, "dtype": None},
     "RFFTN": {"s": None, "dim": None, "norm": None},
     "RFFT": {"n": None, "dim": -1, "norm": None},
+    "RingAttentionUpdate": {"actual_seq_qlen": None, "layout": 'SBH'},
     "RmsNorm": {"epsilon": 1e-6},
     "Roll": {"dims": None},
     "RotaryPositionEmbeddingGrad": {"dx": None, "mode": 0},
@@ -347,6 +345,9 @@ op_args_default_value = {
     "TopKRouter": {"drop_type": 0},
     "TopkExt": {"dim": -1, "largest": True, "sorted": True},
     "TopPRouter": {"drop_type": 0, "threshold": 0.0, "router_prob": 0.0},
+    "ToDevice": {"device": None, "dtype": None, "non_blocking": False, "copy": False},
+    "ToDtype": {"dtype": None, "non_blocking": False, "copy": False},
+    "ToOther": {"non_blocking": False, "copy": False},
     "TraceV2Grad": {"offset": 0, "axis1": 1, "axis2": 0},
     "TraceV2": {"offset": 0, "axis1": 1, "axis2": 0, "dtype": None},
     "TriangularSolve": {"upper": True, "transpose": False, "unitriangular": False},
@@ -381,7 +382,7 @@ op_args_default_value = {
     "FusedInferAttentionScore": {"pse_shift": None, "attn_mask": None, "actual_seq_lengths": None, "actual_seq_lengths_kv": None, "dequant_scale1": None, "quant_scale1": None, "dequant_scale2": None, "quant_scale2": None, "quant_offset2": None, "antiquant_scale": None, "antiquant_offset": None, "block_table": None, "query_padding_size": None, "kv_padding_size": None, "key_antiquant_scale": None, "key_antiquant_offset": None, "value_antiquant_scale": None, "value_antiquant_offset": None, "key_shared_prefix": None, "value_shared_prefix": None, "actual_shared_prefix_len": None, "num_heads": 1, "scale_value": 1.0, "pre_tokens": 2147483647, "next_tokens": 2147483647, "input_layout": 'BSH', "num_key_value_heads": 0, "sparse_mode": 0, "inner_precise": 1, "block_size": 0, "antiquant_mode": 0, "softmax_lse_flag": False, "key_antiquant_mode": 0, "value_antiquant_mode": 0},
     "GroupedMatmul": {"bias": None, "scale": None, "offset": None, "antiquant_scale": None, "antiquant_offset": None, "group_list": None, "split_item": 0, "group_type": -1, "transpose_a": False, "transpose_b": False},
     "GroupedMatmulV2": {"bias": None, "scale": None, "offset": None, "antiquant_scale": None, "antiquant_offset": None, "group_list": None, "split_item": 0, "group_type": -1},
-    "GroupedMatmulV4": {"bias": None, "scale": None, "offset": None, "antiquant_scale": None, "antiquant_offset": None, "pre_token_scale": None, "group_list": None, "activation_input": None, "activation_quant_scale": None, "activation_quant_offset": None, "split_item": 0, "group_type": -1, "group_list_type": 0, "act_type": 0},
+    "GroupedMatmulV4": {"bias": None, "scale": None, "offset": None, "antiquant_scale": None, "antiquant_offset": None, "pre_token_scale": None, "group_list": None, "activation_input": None, "activation_quant_scale": None, "activation_quant_offset": None, "split_item": 0, "group_type": -1, "group_list_type": 0, "act_type": 0, "output_dtype": None},
     "KVCacheScatterUpdate": {"reduce": 'none'},
     "MatmulAllReduceAddRmsNorm": {"reduce_op": 'sum', "comm_turn": 0, "stream_mode": 1},
     "MoeFinalizeRouting": {"x2": None, "bias": None, "scales": None, "expanded_row_idx": None, "expanded_expert_idx": None},
@@ -415,7 +416,6 @@ op_args_default_value = {
     "DeprecatedBaddbmm": {"beta": 1, "alpha": 1},
     "DeprecatedBincount": {"weights": None, "minlength": 0},
     "DeprecatedChunk": {"axis": 0},
-    "DeprecatedInplaceCopy": {"non_blocking": False},
     "DeprecatedCountNonzero": {"axis": (), "keep_dims": False, "dtype": None},
     "DeprecatedCumsum": {"axis": None, "dtype": None},
     "DeprecatedDiv": {"rounding_mode": None},
@@ -430,11 +430,13 @@ op_args_default_value = {
     "DeprecatedMedian": {"axis": -1, "keepdims": False},
     "DeprecatedMin": {"axis": None, "keepdims": False, "initial": None, "where": True, "return_indices": False},
     "DeprecatedNansum": {"axis": None, "keepdims": False, "dtype": None},
+    "DeprecatedPermute": {"axis": None},
     "DeprecatedProd": {"dim": None, "keepdim": False, "dtype": None},
     "DeprecatedRepeatInterleave": {"dim": None},
     "DeprecatedRoll": {"dims": None},
     "DeprecatedSort": {"axis": -1, "descending": False},
     "DeprecatedSplit": {"axis": 0},
+    "DeprecatedSqueeze": {"axis": None},
     "DeprecatedStd": {"axis": None, "ddof": 0, "keepdims": False},
     "DeprecatedSum": {"axis": None, "dtype": None, "keepdims": False, "initial": None},
     "DeprecatedTake": {"axis": None, "mode": 'clip'},
@@ -458,19 +460,10 @@ op_labels = {
     "BroadcastToView": {"side_effect_mem": True},
     "ChunkView": {"side_effect_mem": True},
     "DiagonalView": {"side_effect_mem": True},
-    "DistCommAllReduce": {"side_effect_mem": True},
-    "DistCommReduce": {"side_effect_mem": True},
     "DropoutExt": {"side_effect_hidden": True},
     "DropoutGenMaskExt": {"side_effect_hidden": True},
     "Dropout": {"side_effect_hidden": True},
-    "EmbeddingApplyAdamW": {"_process_node_engine_id": 'PS'},
-    "EmbeddingApplyAdam": {"_process_node_engine_id": 'PS'},
-    "EmbeddingApplyAdaGrad": {"_process_node_engine_id": 'PS'},
-    "EmbeddingApplyFtrl": {"_process_node_engine_id": 'PS'},
-    "EmbeddingApplyRmsprop": {"_process_node_engine_id": 'PS'},
-    "EmbeddingApplySgd": {"_process_node_engine_id": 'PS'},
     "Embedding": {"side_effect_mem": True},
-    "EmbeddingTableEvict": {"_process_node_engine_id": 'PS'},
     "ExpandDimsView": {"side_effect_mem": True},
     "Generator": {"side_effect_mem": True},
     "GroupTopk": {"side_effect_mem": True},
@@ -478,6 +471,8 @@ op_labels = {
     "InplaceAddmm": {"side_effect_mem": True},
     "InplaceAddsExt": {"side_effect_mem": True},
     "InplaceAddExt": {"side_effect_mem": True},
+    "InplaceBernoulliScalar": {"side_effect_mem": True},
+    "InplaceBernoulliTensor": {"side_effect_mem": True},
     "InplaceCopy": {"side_effect_mem": True},
     "InplaceDivMods": {"side_effect_mem": True},
     "InplaceDivMod": {"side_effect_mem": True},
@@ -495,10 +490,12 @@ op_labels = {
     "InplaceGroupedMatmulAdd": {"side_effect_mem": True},
     "InplaceHardtanh": {"side_effect_mem": True},
     "InplaceIndexAddExt": {"side_effect_mem": True},
+    "InplaceIndexCopy": {"side_effect_mem": True},
     "InplaceIndexPut": {"side_effect_mem": True},
     "InplaceLog": {"side_effect_mem": True},
     "InplaceMaskedFillScalar": {"side_effect_mem": True},
     "InplaceMaskedFillTensor": {"side_effect_mem": True},
+    "InplaceMaskedScatter": {"side_effect_mem": True},
     "InplaceMatmulAdd": {"side_effect_mem": True},
     "InplaceMuls": {"side_effect_mem": True},
     "InplaceMul": {"side_effect_mem": True},
@@ -506,18 +503,24 @@ op_labels = {
     "InplacePut": {"side_effect_mem": True},
     "InplaceRandom": {"side_effect_mem": True},
     "InplaceReLU": {"side_effect_mem": True},
+    "InplaceRemainderTensorScalar": {"side_effect_mem": True},
+    "InplaceRemainderTensorTensor": {"side_effect_mem": True},
     "InplaceScatterAdd": {"side_effect_mem": True},
     "InplaceScatterSrc": {"side_effect_mem": True},
     "InplaceScatterSrcReduce": {"side_effect_mem": True},
     "InplaceScatterValue": {"side_effect_mem": True},
     "InplaceScatterValueReduce": {"side_effect_mem": True},
+    "InplaceSigmoid": {"side_effect_mem": True},
+    "InplaceSign": {"side_effect_mem": True},
     "InplaceSiLU": {"side_effect_mem": True},
     "InplaceSubExt": {"side_effect_mem": True},
     "InplaceSubScalar": {"side_effect_mem": True},
     "InplaceTanh": {"side_effect_mem": True},
     "InplaceThreshold": {"side_effect_mem": True},
     "InplaceUniform": {"side_effect_mem": True},
+    "KvScaleCache": {"side_effect_mem": True},
     "Log": {"cust_aicpu": 'Log', "base": -1.0, "scale": 1.0, "shift": 0.0},
+    "MlaPreprocess": {"side_effect_mem": True},
     "NarrowView": {"side_effect_mem": True},
     "ReshapeAndCache": {"side_effect_mem": True},
     "ResizeD": {"mode": 'linear'},
@@ -531,6 +534,7 @@ op_labels = {
     "TransposeView": {"side_effect_mem": True},
     "UnstackExtView": {"side_effect_mem": True},
     "KVCacheScatterUpdate": {"side_effect_mem": True},
+    "DistCommAllReduce": {"side_effect_mem": True},
+    "DistCommReduce": {"side_effect_mem": True},
     "InplaceExponential": {"side_effect_mem": True},
-    "DeprecatedInplaceCopy": {"side_effect_mem": True},
 }

mindspore/ops/auto_generate/gen_extend_func.py

@@ -118,11 +118,11 @@ def add(input, other, alpha=1):
         input (Union[Tensor, number.Number, bool]): The first input is a number.Number or
             a bool or a tensor whose data type is
             `number <https://www.mindspore.cn/docs/en/master/api_python/mindspore/mindspore.dtype.html>`_ or
-            `bool_ <https://www.mindspore.cn/docs/en/master/api_python/mindspore/mindspore.dtype.html>`_.
+            `bool <https://www.mindspore.cn/docs/en/master/api_python/mindspore/mindspore.dtype.html>`_.
         other (Union[Tensor, number.Number, bool]): The second input, is a number.Number or
             a bool or a tensor whose data type is
             `number <https://www.mindspore.cn/docs/en/master/api_python/mindspore/mindspore.dtype.html>`_ or
-            `bool_ <https://www.mindspore.cn/docs/en/master/api_python/mindspore/mindspore.dtype.html>`_.
+            `bool <https://www.mindspore.cn/docs/en/master/api_python/mindspore/mindspore.dtype.html>`_.
         alpha (number.Number): A scaling factor applied to `other`, default 1.
     
     Returns:
@@ -905,7 +905,6 @@ def histc(input, bins=100, min=0, max=0):
     Elements lower than min or higher than max are ignored.
     
     .. warning::
-        This is an experimental API that is subject to change or deletion.
         If input is int64, valid values fit within int32; exceeding this may cause precision errors.
     
     Args:
@@ -1732,9 +1731,6 @@ def outer(input, vec2):
     Return outer product of `input` and `vec2`. If `input` is a vector of size :math:`n`
     and `vec2` is a vector of size :math:`m` , then output must be a matrix of shape :math:`(n, m)` .
     
-    .. warning::
-        This is an experimental API that is subject to change or deletion.
-    
     .. note::
         This function does not broadcast.
     
@@ -1957,14 +1953,13 @@ def stack(tensors, dim=0):
     :math:`(x_1, x_2, ..., x_{dim}, N, x_{dim+1}, ..., x_R)`.
     
     Args:
-        tensors (Union[tuple, list]): A Tuple or list of Tensor objects with the same shape and type.
+        tensors (Union[tuple, list]): A Tuple or list of Tensor objects with the same shape.
         dim (int, optional): Dimension to stack. The range is [-(R+1), R+1). Default: ``0`` .
     
     Returns:
-        Tensor. A stacked Tensor with the same type as `tensors`.
+        A stacked Tensor.
     
     Raises:
-        TypeError: If the data types of elements in `tensors` are not the same.
         ValueError: If `dim` is out of the range [-(R+1), R+1);
                     or if the shapes of elements in `tensors` are not the same.
     
@@ -2003,11 +1998,11 @@ def sub(input, other, alpha=1):
         input (Union[Tensor, number.Number, bool]): The first input is a number.Number or
             a bool or a tensor whose data type is
             `number <https://www.mindspore.cn/docs/en/master/api_python/mindspore/mindspore.dtype.html>`_ or
-            `bool_ <https://www.mindspore.cn/docs/en/master/api_python/mindspore/mindspore.dtype.html>`_.
+            `bool <https://www.mindspore.cn/docs/en/master/api_python/mindspore/mindspore.dtype.html>`_.
         other (Union[Tensor, number.Number, bool]): The second input, is a number.Number or
             a bool or a tensor whose data type is
             `number <https://www.mindspore.cn/docs/en/master/api_python/mindspore/mindspore.dtype.html>`_ or
-            `bool_ <https://www.mindspore.cn/docs/en/master/api_python/mindspore/mindspore.dtype.html>`_.
+            `bool <https://www.mindspore.cn/docs/en/master/api_python/mindspore/mindspore.dtype.html>`_.
         alpha (number.Number): A scaling factor applied to `other`, default 1.
     
     Returns: