mindstudio-probe 1.3.0__py3-none-any.whl → 8.1.1__py3-none-any.whl

msprobe/README.md

@@ -54,7 +54,9 @@ export MSPROBE_LOG_LEVEL={x}
 
 **2. 工具读写的所有路径，如config_path、dump_path等，只允许包含大小写字母、数字、下划线、斜杠、点和短横线。**
 
-## ⚙️ [安装](./docs/01.installation.md)
+## ⚙️ 安装
+
+请参见[安装指导说明](./docs/01.installation.md)。
 
 ## 🌟 新版本特性
 
@@ -138,6 +140,8 @@ MindSpore 动态图场景的[离线预检](./docs/09.accuracy_checker_MindSpore.
 
 [PyTorch 单算子API自动生成脚本](./docs/23.generate_operator_PyTorch.md)
 
+[MindSpore 单算子API自动生成脚本](./docs/33.generate_operator_MindSpore.md)
+
 ### 11 数码关联
 
 该功能只支持 MindSpore 静态图场景，用于将IR图与dump数据进行关联，获取dump数据和代码调用栈的关联关系。
@@ -155,6 +159,33 @@ MindSpore 动态图场景的[离线预检](./docs/09.accuracy_checker_MindSpore.
 
 [MSAdapter 场景的溢出检测](./docs/30.overflow_check_MSAdapter.md)
 
+### 13 训练检查
+
+该工具主要包括：
+
+训练前或精度比对前，对比两个环境下可能影响训练精度的配置差异。
+
+[PyTorch 训练前配置检查](./docs/31.config_check.md)
+
+训练过程中或结束后，比较两个不同的checkpoint，评估模型相似度。
+
+[checkpoint比对](./docs/32.ckpt_compare.md)
+
+### 14 强化学习数据采集
+
+主要能力：
+
+灵活采集强化学习中重要关键过程数据，并支持比对。
+
+[强化学习数据采集](./docs/34.RL_collect.md)
+
+### 15 整网首个溢出节点分析
+
+多rank场景下通过dump数据找到首个出现Nan或Inf的节点。
+
+[PyTorch 场景整网首个溢出节点分析](./docs/35.nan_analyze.md)
+
+
 ## 📑 补充材料
 
 [无标杆比对功能在 PyTorch 场景的性能基线报告](./docs/S02.report_free_benchmarking_validation_performance_baseline.md)

msprobe/core/__init__.py

@@ -0,0 +1,17 @@
+# Copyright (c) 2025-2025, Huawei Technologies Co., Ltd.
+# All rights reserved.
+#
+# 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.
+
+from msprobe.core.single_save.single_saver import SingleSave
+from msprobe.core.single_save.single_comparator import SingleComparator

msprobe/core/common/const.py

@@ -70,7 +70,7 @@ class Const:
     SUMMARY = "summary"
     MD5 = "md5"
     VALUE = "value"
-    SUMMARY_MODE = [ALL, SUMMARY, MD5]
+    SUMMARY_MODE = ["statistics", "md5"]
 
     WRITE_FLAGS = os.O_WRONLY | os.O_CREAT
     WRITE_MODES = stat.S_IWUSR | stat.S_IRUSR
@@ -80,6 +80,8 @@ class Const:
     NUMPY_SUFFIX = ".npy"
     NUMPY_PATTERN = "*.npy"
     PT_SUFFIX = ".pt"
+    PY_SUFFIX = ".py"
+    INIT_PY = "init.py"
     ONE_GB = 1073741824  # 1 * 1024 * 1024 * 1024
     TEN_GB = 10737418240  # 10 * 1024 * 1024 * 1024
     ONE_MB = 1048576  # 1 * 1024 * 1024
@@ -95,6 +97,7 @@ class Const:
     GRAD_OUTPUT = 'grad_output'
     PARAMS = 'parameters'
     PARAMS_GRAD = 'parameters_grad'
+    DEBUG = 'debug'
     START = "start"
     STOP = "stop"
     ENV_ENABLE = "1"
@@ -132,6 +135,7 @@ class Const:
     NPU = 'NPU'
     NPU_LOWERCASE = 'npu'
     CPU_LOWERCASE = 'cpu'
+    GPU_LOWERCASE = 'gpu'
     CUDA_LOWERCASE = 'cuda'
     DEVICE = 'device'
     DISTRIBUTED = 'Distributed'
@@ -140,6 +144,10 @@ class Const:
     MODULE_PREFIX = ["Module", "Cell"]
     FORWARD_NAME_SUFFIX = ".forward"
 
+    DUMP_JSON_FILE = "dump_json_file"
+    DEBUG_JSON_FILE = "debug_json_file"
+    STACK_JSON_FILE = "stack_json_file"
+
     # struct json param
     ORIGIN_DATA = "origin_data"
     SCOPE = "scope"
@@ -170,6 +178,10 @@ class Const:
     TOP_LAYER = "TopLayer"
     CELL = "Cell"
     MODULE = "Module"
+    API = "api"
+    PYNATIVE_MODE = "pynative"
+    PYNATIVE_GRAPH_MODE = "pynative_graph"
+
     FRAME_FILE_LIST = ["site-packages/torch", "package/torch", "site-packages/mindspore", "package/mindspore"]
     INPLACE_LIST = [
         "broadcast", "all_reduce", "reduce", "all_gather", "gather", "scatter", "reduce_scatter",
@@ -191,7 +203,11 @@ class Const:
 
     FILL_CHAR_NUMS = 50
     TOOL_ENDS_SUCCESSFULLY = f"{TOOL_NAME} ends successfully."
+
     WITHOUT_CALL_STACK = "The call stack retrieval failed."
+    STACK_FILTER_KEYWORDS = ["msprobe/core", "msprobe/pytorch", "msprobe/mindspore"]
+    CALL_STACK_FLAG = "data_dump/api_registry"
+    NEW_STACK_FLAG = "0"
 
     STEP = "step"
     RANK = "rank"
@@ -209,12 +225,16 @@ class Const:
     TORCH_FLOAT32 = "torch.float32"
     TORCH_BFLOAT16 = "torch.bfloat16"
 
+    TYPE = 'type'
     DTYPE = 'dtype'
     SHAPE = 'shape'
+    STACK_INFO = 'stack_info'
     MAX = 'Max'
     MIN = 'Min'
     MEAN = 'Mean'
     NORM = 'Norm'
+    DATA_NAME = 'data_name'
+    TENSOR_STAT_INDEX = 'tensor_stat_index'
 
     CODE_STACK = 'Code Stack'
     OP_NAME = 'Op Name'
@@ -226,6 +246,10 @@ class Const:
     # 分隔符常量
     SCOPE_SEPARATOR = "/"
     REPLACEMENT_CHARACTER = "_"
+    PIPE_SEPARATOR = "|"
+
+    FORWARD_PATTERN = SEP + FORWARD + SEP
+    BACKWARD_PATTERN = SEP + BACKWARD + SEP
 
     OPTIMIZER = "optimizer"
     CLIP_GRAD = "clip_grad"
@@ -243,6 +267,7 @@ class Const:
     PT_API_TYPE_ATEN = "aten"
     PT_API_TYPE_DIST = "distributed"
     PT_API_TYPE_NPU_DIST = "npu_distributed"
+    PT_API_TYPE_MINDSPEED = "mindspeed"
 
     MS_API_TYPE_OPS = "ops"
     MS_API_TYPE_TENSOR = "tensor"
@@ -250,6 +275,7 @@ class Const:
     MS_API_TYPE_MINT = "mint.ops"
     MS_API_TYPE_MINT_FUNC = "mint.nn.functional"
     MS_API_TYPE_COM = "communication.comm_func"
+    MS_API_TYPE_MINT_DIST = "mint.distributed"
 
     FUNCTIONAL_API_TYPE_PREFIX = "Functional"
     TENSOR_API_TYPE_PREFIX = "Tensor"
@@ -259,9 +285,11 @@ class Const:
     NPU_API_TYPE_PREFIX = "NPU"
     ATEN_API_TYPE_PREFIX = "Aten"
     VF_API_TYPE_PREFIX = "VF"
+    MINDSPEED_API_TYPE_PREFIX = "MindSpeed"
 
     MINT_API_TYPE_PREFIX = "Mint"
     MINT_FUNC_API_TYPE_PREFIX = "MintFunctional"
+    MINT_DIST_API_TYPE_PREFIX = "MintDistributed"
 
     SUPPORT_API_DICT_KEY_MAP = {
         PT_FRAMEWORK: {
@@ -272,7 +300,8 @@ class Const:
             PT_API_TYPE_NPU: PT_API_TYPE_NPU,
             PT_API_TYPE_ATEN: PT_API_TYPE_ATEN,
             PT_API_TYPE_DIST: PT_API_TYPE_DIST,
-            PT_API_TYPE_NPU_DIST: PT_API_TYPE_NPU_DIST
+            PT_API_TYPE_NPU_DIST: PT_API_TYPE_NPU_DIST,
+            PT_API_TYPE_MINDSPEED: PT_API_TYPE_MINDSPEED
         },
         MS_FRAMEWORK: {
             MS_API_TYPE_OPS: MS_API_TYPE_OPS,
@@ -280,7 +309,8 @@ class Const:
             MS_API_TYPE_STUB_TENSOR: MS_API_TYPE_TENSOR,
             MS_API_TYPE_MINT: MS_API_TYPE_MINT,
             MS_API_TYPE_MINT_FUNC: MS_API_TYPE_MINT_FUNC,
-            MS_API_TYPE_COM: MS_API_TYPE_COM
+            MS_API_TYPE_COM: MS_API_TYPE_COM,
+            MS_API_TYPE_MINT_DIST: MS_API_TYPE_MINT_DIST
         },
         MT_FRAMEWORK: {
             PT_API_TYPE_FUNCTIONAL: PT_API_TYPE_FUNCTIONAL,
@@ -300,7 +330,8 @@ class Const:
             PT_API_TYPE_NPU: NPU_API_TYPE_PREFIX,
             PT_API_TYPE_ATEN: ATEN_API_TYPE_PREFIX,
             PT_API_TYPE_DIST: DIST_API_TYPE_PREFIX,
-            PT_API_TYPE_NPU_DIST: DIST_API_TYPE_PREFIX
+            PT_API_TYPE_NPU_DIST: DIST_API_TYPE_PREFIX,
+            PT_API_TYPE_MINDSPEED: MINDSPEED_API_TYPE_PREFIX
         },
         MS_FRAMEWORK: {
             MS_API_TYPE_OPS: FUNCTIONAL_API_TYPE_PREFIX,
@@ -308,7 +339,8 @@ class Const:
             MS_API_TYPE_STUB_TENSOR: TENSOR_API_TYPE_PREFIX,
             MS_API_TYPE_MINT: MINT_API_TYPE_PREFIX,
             MS_API_TYPE_MINT_FUNC: MINT_FUNC_API_TYPE_PREFIX,
-            MS_API_TYPE_COM: DIST_API_TYPE_PREFIX
+            MS_API_TYPE_COM: DIST_API_TYPE_PREFIX,
+            MS_API_TYPE_MINT_DIST: MINT_DIST_API_TYPE_PREFIX
         },
         MT_FRAMEWORK: {
             PT_API_TYPE_FUNCTIONAL: FUNCTIONAL_API_TYPE_PREFIX,
@@ -319,12 +351,42 @@ class Const:
         }
     }
 
+    def _fused_adamw_(
+        self,
+        grads,
+        exp_avgs,
+        exp_avg_sqs,
+        max_exp_avg_sqs,
+        state_steps,
+        *,
+        lr,
+        beta1,
+        beta2,
+        weight_decay,
+        eps,
+        amsgrad,
+        maximize,
+        grad_scale=None,
+        found_inf=None
+    ):
+        pass
+
+    API_WITH_SELF_ARG = {
+        'Torch._fused_adamw_': _fused_adamw_
+    }
+
+    ASCEND = "ASCEND"
+    MATCH_MODE_NAME = "pure name"
+    MATCH_MODE_MAPPING = "mapping"
+    MATCH_MODE_SIMILARITY = "similarity"
+
 
 class CompareConst:
     """
     Class for compare module const
     """
     SPACE = " "
+    NAME = "Name"
     # compare result column name
     NPU_NAME = "NPU Name"
     BENCH_NAME = "Bench Name"
@@ -368,6 +430,7 @@ class CompareConst:
     OUTPUT_STRUCT = "output_struct"
     PARAMS_STRUCT = "params_struct"
     PARAMS_GRAD_STRUCT = "params_grad_struct"
+    DEBUG_STRUCT = "debug_struct"
     SUMMARY = "summary"
     COMPARE_RESULT = "compare_result"
     COMPARE_MESSAGE = "compare_message"
@@ -474,16 +537,10 @@ class CompareConst:
         Const.KWARGS: INPUT_STRUCT,
         Const.OUTPUT: OUTPUT_STRUCT,
         Const.PARAMS: PARAMS_STRUCT,
-        Const.PARAMS_GRAD: PARAMS_GRAD_STRUCT
+        Const.PARAMS_GRAD: PARAMS_GRAD_STRUCT,
+        Const.DEBUG: DEBUG_STRUCT
     }
 
-    STRUCT_COMPARE_KEY = [
-        INPUT_STRUCT,
-        OUTPUT_STRUCT,
-        PARAMS_STRUCT,
-        PARAMS_GRAD_STRUCT
-    ]
-
     # compare standard
     HUNDRED_RATIO_THRESHOLD = 0.01
     THOUSAND_RATIO_THRESHOLD = 0.001
@@ -562,15 +619,35 @@ class CompareConst:
         MAX_DIFF: None, MIN_DIFF: None, MEAN_DIFF: None, NORM_DIFF: None, MAX_RELATIVE_ERR: None,
         MIN_RELATIVE_ERR: None, MEAN_RELATIVE_ERR: None, NORM_RELATIVE_ERR: None
     }
+
+    API_MAPPING_KEYS_TO_COMPARE = [
+        ('ms_args', 'pt_args'),
+        ('ms_outputs', 'pt_outputs'),
+        ('ms_parameters', 'pt_parameters'),
+        ('ms_parameters_grad', 'pt_parameters_grad')
+    ]
+
     INPUT_PATTERN = Const.SEP + Const.INPUT + Const.SEP
     KWARGS_PATTERN = Const.SEP + Const.KWARGS + Const.SEP
     OUTPUT_PATTERN = Const.SEP + Const.OUTPUT + Const.SEP
     PARAMS_PATTERN = Const.SEP + Const.PARAMS + Const.SEP
     PARAMS_GRAD_PATTERN = Const.SEP + Const.PARAMS_GRAD + Const.SEP
-    COMPARE_KEY = 'compare_key'
-    COMPARE_SHAPE = 'compare_shape'
+
+    CMP_KEY = 'compare_key'
+    CMP_SHAPE = 'compare_shape'
+
+    OP_NAME_X = 'op_name_x'
+    MATCH_RESULT_COLUMNS = [
+        OP_NAME_X, 'dtype_x', 'shape_x', 'summary_x', 'stack_info_x', 'data_name_x',
+        CMP_KEY, CMP_SHAPE,
+        'op_name_y', 'dtype_y', 'shape_y', 'summary_y', 'stack_info_y', 'data_name_y',
+    ]
+
     INTERNAL_API_MAPPING_FILE = 'ms_to_pt_api.yaml'
     UNREADABLE = 'unreadable data'
+    NPU_DUMP_DATA_DIR = 'npu_dump_data_dir'
+    BENCH_DUMP_DATA_DIR = 'bench_dump_data_dir'
+    NO_REAL_DATA_FLAG = '-1'
 
 
 class FileCheckConst:
@@ -592,6 +669,8 @@ class FileCheckConst:
     XLSX_SUFFIX = ".xlsx"
     YAML_SUFFIX = ".yaml"
     IR_SUFFIX = ".ir"
+    ZIP_SUFFIX = ".zip"
+    SHELL_SUFFIX = ".sh"
     MAX_PKL_SIZE = 1073741824  # 1 * 1024 * 1024 * 1024
     MAX_NUMPY_SIZE = 10737418240  # 10 * 1024 * 1024 * 1024
     MAX_JSON_SIZE = 1073741824  # 1 * 1024 * 1024 * 1024
@@ -600,6 +679,9 @@ class FileCheckConst:
     MAX_XLSX_SIZE = 1073741824  # 1 * 1024 * 1024 * 1024
     MAX_YAML_SIZE = 1073741824  # 1 * 1024 * 1024 * 1024
     MAX_IR_SIZE = 1073741824  # 1 * 1024 * 1024 * 1024
+    MAX_ZIP_SIZE = 10737418240  # 10 * 1024 * 1024 * 1024
+    MAX_FILE_IN_ZIP_SIZE = 1073741824  # 1 * 1024 * 1024 * 1024
+    MAX_FILE_SIZE = 1073741824  # 1 * 1024 * 1024 * 1024
     COMMOM_FILE_SIZE = 1048576  # 1 * 1024 * 1024
     DIR = "dir"
     FILE = "file"
@@ -613,7 +695,8 @@ class FileCheckConst:
         CSV_SUFFIX: MAX_CSV_SIZE,
         XLSX_SUFFIX: MAX_XLSX_SIZE,
         YAML_SUFFIX: MAX_YAML_SIZE,
-        IR_SUFFIX: MAX_IR_SIZE
+        IR_SUFFIX: MAX_IR_SIZE,
+        ZIP_SUFFIX: MAX_ZIP_SIZE
     }
     CSV_BLACK_LIST = r'^[＋－＝％＠\+\-=%@]|;[＋－＝％＠\+\-=%@]'
 
@@ -671,7 +754,7 @@ class MonitorConst:
     DEFAULT_MIN_COLLECT_TIMES = 0
     DEFAULT_STEP_INTERVAL = 1
 
-    OP_LIST = ["norm", "min", "max", "zeros", "nans", "id", "mean"]
+    OP_LIST = ["norm", "min", "max", "zeros", "nans", "id", "mean", "shape", "dtype"]
     MONITOR_OUTPUT_DIR = "MONITOR_OUTPUT_DIR"
     DEFAULT_MONITOR_OUTPUT_DIR = "./monitor_output"
     DATABASE = "database"
@@ -683,7 +766,7 @@ class MonitorConst:
         "DeepSpeedZeroOptimizer_Stage3"
     )
     DEEPSPEED_ZERO_OPT_FILTER = "DeepSpeedZeroOptimizer"
-    RULE_NAME = ['AnomalyTurbulence']
+    RULE_NAME = ['AnomalyTurbulence', 'AnomalyNan']
 
     SLICE_SIZE = 20480
     # used for name
@@ -700,15 +783,16 @@ class MonitorConst:
     ACTVGRAD = "actv_grad"
     POST_GRAD = "post_grad"
     PRE_GRAD = "pre_grad"
+    PRE_PARAM = "param_origin"
+    POST_PARAM = "param_updated"
     ACC_GRAD = "acc_grad"
     PREFIX_POST = "post"
     PREFIX_PRE = "pre"
     EXP_AVG = "exp_avg"
     EXP_AVG_SQ = "exp_avg_sq"
-    PARAM = "param"
 
     CSV_HEADER = ["vpp_stage", "name", "step"]
-    CSV_HEADER_XY = ["vpp_stage", "name", "step", "micro_step"]
+    CSV_HEADER_MICRO_STEP = ["vpp_stage", "name", "step", "micro_step"]
     OUTPUT_DIR_PATTERN = r"([\w-]{0,20})-rank(\d{1,5})-"
     ANOMALY_JSON = "anomaly.json"
     ANALYSE_JSON = "anomaly_analyse.json"
@@ -716,5 +800,20 @@ class MonitorConst:
     CSV = "csv"
     API = "api"
     HEADER_NAME = 'name'
-
     MAX_NDIGITS = 20
+
+    DEFAULT_STAGE = -1
+    FORWARD_STAGE = 0
+    BACKWARD_STAGE = 1
+    OPTIMIZER_STAGE = 2
+    FORWARD_KEY = [ACTV]
+    BACKWARD_KEY = [ACTVGRAD, PRE_GRAD, POST_GRAD, ACC_GRAD]
+    OPTIMIZER_KEY = [EXP_AVG, EXP_AVG_SQ]
+
+    TRAIN_STAGE = {}
+    for key in FORWARD_KEY:
+        TRAIN_STAGE[key] = FORWARD_STAGE
+    for key in BACKWARD_KEY:
+        TRAIN_STAGE[key] = BACKWARD_STAGE
+    for key in OPTIMIZER_KEY:
+        TRAIN_STAGE[key] = OPTIMIZER_STAGE

msprobe/core/common/exceptions.py

@@ -21,8 +21,8 @@ class CodedException(Exception):
 
     def __str__(self):
         return self.error_info
-
-
+    
+    
 class MsprobeException(CodedException):
     INVALID_PARAM_ERROR = 0
     OVERFLOW_NUMS_ERROR = 1