diff --git a/.gitignore b/.gitignore
index a81c8ee121952cf06bfaf9ff9988edd8cded763c..c70c40e0f527c8c20a6bf994bcb8070b95e13e27 100644
--- a/.gitignore
+++ b/.gitignore
@@ -136,3 +136,10 @@ dmypy.json
 
 # Cython debug symbols
 cython_debug/
+
+# vscode settings and analysis output
+.vscode/
+att_advisor*.html
+*.xlsx
+operator_tuning_file*.cfg
+.ipynb_checkpoints/
\ No newline at end of file
diff --git a/profiler/advisor/analyzer/base_analyzer.py b/profiler/advisor/analyzer/base_analyzer.py
index 7e0ce3b8968f4cfe64da3ad33eb09eb1e6e50ab0..5f4bd3202cd2071088f25564a7d4b14144a34826 100644
--- a/profiler/advisor/analyzer/base_analyzer.py
+++ b/profiler/advisor/analyzer/base_analyzer.py
@@ -17,11 +17,10 @@ class BaseAnalyzer(VersionControl, metaclass=ABCMeta):
 
     dataset_cls_list = []
 
-    def __init__(self, collection_path, n_processes: int = 1, cann_version=constant.DEFAULT_CANN_VERSION,
-                 torch_version=constant.DEFAULT_TORCH_VERSION, **kwargs):
+    def __init__(self, collection_path, n_processes: int = 1, **kwargs):
         self.n_processes = n_processes
-        self.cann_version = cann_version
-        self.torch_version = torch_version
+        self.cann_version = kwargs.get("cann_version", constant.DEFAULT_CANN_VERSION)
+        self.torch_version = kwargs.get("torch_version", constant.DEFAULT_TORCH_VERSION)
         self.html_render = HTMLRender()
         self.collection_path = collection_path
         self.kwargs = kwargs
@@ -41,7 +40,7 @@ class BaseAnalyzer(VersionControl, metaclass=ABCMeta):
         def decorate(func):
 
             @wraps(func)
-            def wrapper(self):
+            def wrapper(self, **kwargs):
                 data = self.dataset_list
                 if data is None:
                     return None
@@ -57,7 +56,7 @@ class BaseAnalyzer(VersionControl, metaclass=ABCMeta):
         return decorate
 
     @abstractmethod
-    def optimize(self):
+    def optimize(self, **kwargs):
         pass
 
     @abstractmethod
diff --git a/profiler/advisor/analyzer/cluster/slow_link_analyser.py b/profiler/advisor/analyzer/cluster/slow_link_analyser.py
index a4c38187a792d6bde3ed5426b62b165c87627acf..846b79a50f31abb8445a0e5c2e82aaaf3c8ee23d 100644
--- a/profiler/advisor/analyzer/cluster/slow_link_analyser.py
+++ b/profiler/advisor/analyzer/cluster/slow_link_analyser.py
@@ -37,9 +37,8 @@ class SlowLinkAnalyzer(BaseAnalyzer):
     SLOW_LINK_ANALYSIS = "slow_link_analysis"
     dataset_cls_list = [ClusterCommunicationDataSet]
 
-    def __init__(self, collection_path, n_processes: int = 1, cann_version=constant.DEFAULT_CANN_VERSION,
-                 torch_version=constant.DEFAULT_TORCH_VERSION, **kwargs):
-        super().__init__(collection_path, n_processes, cann_version, torch_version, **kwargs)
+    def __init__(self, collection_path, n_processes: int = 1, **kwargs):
+        super().__init__(collection_path, n_processes, **kwargs)
         key = ClusterCommunicationDataSet.get_key()
         self.communication_data_class = self.get_first_data_by_key(self.dataset_list, key)
         self.rank_bw_dict = self.communication_data_class.get_data()
diff --git a/profiler/advisor/analyzer/cluster/slow_rank_analyser.py b/profiler/advisor/analyzer/cluster/slow_rank_analyser.py
index ff80178e239feed79809c72e306356b9e305017e..4215b514a215a2a350571746ff9cb90c3c9956eb 100644
--- a/profiler/advisor/analyzer/cluster/slow_rank_analyser.py
+++ b/profiler/advisor/analyzer/cluster/slow_rank_analyser.py
@@ -29,9 +29,8 @@ class SlowRankAnalyzer(BaseAnalyzer):
     BOTTLENECK_LIST = ['Computing', 'Communication', "Free"]
     dataset_cls_list = [ClusterStepTraceTimeDataSet]
 
-    def __init__(self, collection_path, n_processes: int = 1, cann_version=constant.DEFAULT_CANN_VERSION,
-                 torch_version=constant.DEFAULT_TORCH_VERSION, **kwargs):
-        super().__init__(collection_path, n_processes, cann_version, torch_version, **kwargs)
+    def __init__(self, collection_path, n_processes: int = 1, **kwargs):
+        super().__init__(collection_path, n_processes, **kwargs)
         key = ClusterStepTraceTimeDataSet.get_key()
         self.step_trace_class =  self.get_first_data_by_key(self.dataset_list, key)
         self.step_trace_dict = self.step_trace_class.get_data()
diff --git a/profiler/advisor/analyzer/computation/bound/block_dim_checker.py b/profiler/advisor/analyzer/computation/bound/block_dim_checker.py
index d90ef56c78b52987bc7d1dbb2c17d5caf9eb7706..a7d7ddd93c70e59dc0d10318fdac06fdc581f70c 100644
--- a/profiler/advisor/analyzer/computation/bound/block_dim_checker.py
+++ b/profiler/advisor/analyzer/computation/bound/block_dim_checker.py
@@ -20,7 +20,11 @@ class BlockDimChecker(OperatorChecker):
         "input_data_types", "input_formats", "output_shapes", "output_data_types", "output_formats"
     ]
 
+    def pre_check(self, profiling_data) -> bool:
+        return not self.is_dynamic_shape(profiling_data)
+
     def _check_data(self, data):
+        self.format_suggestion_content(data)
         if not self._check_summary(data):
             return False
         if not Config().get_config("ai_core_num"):
@@ -69,9 +73,3 @@ class BlockDimChecker(OperatorChecker):
         else:
             core_num = self._aiv_num
         return core_num
-
-    def format_suggestion_content(self, profiling_data: ProfilingDataset) -> None:
-        if profiling_data.PROF_TYPE == constant.ASCEND_PYTORCH_PROFILER:
-            self._SUGGESTION.append(self.PyTorch_OPERATOR_TUNE_SUGGESTION)
-        elif profiling_data.PROF_TYPE == constant.MSLITE:
-            self._SUGGESTION.append(self.MSLite_OPERATOR_TUNE_SUGGESTION)
diff --git a/profiler/advisor/analyzer/computation/bound/operator_bound_checker.py b/profiler/advisor/analyzer/computation/bound/operator_bound_checker.py
index 4ede3c94e6703f2ce38f9339db8fe9405fcfa82f..a22b380f974b14207d6d7be262cd49f0ba0fbe99 100644
--- a/profiler/advisor/analyzer/computation/bound/operator_bound_checker.py
+++ b/profiler/advisor/analyzer/computation/bound/operator_bound_checker.py
@@ -24,12 +24,15 @@ class OperatorBoundChecker(OperatorChecker):
         "output_data_types", "output_formats"
     ]
 
+    def pre_check(self, profiling_data) -> bool:
+        return not self.is_dynamic_shape(profiling_data)
+
     def _check_data(self, data):
+        self.format_suggestion_content(data)
         if not self._check_summary(data):
             return False
         for op_info in data.op_summary.op_list:
-            if self._check_operator(op_info):
-                return True
+            return self._check_operator(op_info)
 
         logger.warning(self.SKIP_CHECK_MSG, self._CHECKER, "ratio in op summary")
         return False
@@ -48,9 +51,3 @@ class OperatorBoundChecker(OperatorChecker):
                                     template_dir="templates",
                                     template_name="operator_no_bound.html",
                                     format_result=self.format_operator_result(record, constant.OPERATOR_OUT_TOPK))
-
-    def format_suggestion_content(self, profiling_data: ProfilingDataset) -> None:
-        if profiling_data.PROF_TYPE == constant.ASCEND_PYTORCH_PROFILER:
-            self._SUGGESTION.append(self.PyTorch_OPERATOR_TUNE_SUGGESTION)
-        elif profiling_data.PROF_TYPE == constant.MSLITE:
-            self._SUGGESTION.append(self.MSLite_OPERATOR_TUNE_SUGGESTION)
diff --git a/profiler/advisor/analyzer/computation/op_compile/dynamic_shape_checker.py b/profiler/advisor/analyzer/computation/op_compile/dynamic_shape_checker.py
index 070b3a3b57b0a3d850a2e34bf408c5cf6c2a9610..86d3bac4ff8cb163d23a6365307b855839b12a6a 100644
--- a/profiler/advisor/analyzer/computation/op_compile/dynamic_shape_checker.py
+++ b/profiler/advisor/analyzer/computation/op_compile/dynamic_shape_checker.py
@@ -25,33 +25,7 @@ class DynamicShapeChecker(OperatorChecker):
         super().__init__(cann_version=cann_version)
 
     def check(self, profiling_database) -> bool:
-        less_than_cann800_list = [constant.CANN_VERSION_C30, constant.CANN_VERSION_C13, constant.CANN_VERSION_C15]
-        # CANN 8.0.0 之前从 ge_info 中获取 op_state 属性，进行动态 shape 逻辑判断
-        if self.cann_version in less_than_cann800_list:
-            if hasattr(profiling_database, "ge_info"):
-                ge_info = profiling_database.ge_info
-                static_shape_operators = ge_info.get_static_shape_operators()
-                if len(static_shape_operators) == 0:
-                    OperatorChecker.IS_ALL_OPERATOR_DYNAMIC_SHAPE = True
-                    return True
-            else:
-                logger.warning(
-                    "Skip dynamic shape checker because of not containing ge_info.db file in host filefloder.\n"
-                    "To enable dynamic shape checker, please try to set data_simplification=False in experimental_config.\n"
-                    "More details please refer to link : %s", constant.ASCEND_PROFILER_URL)
-        else:
-            # CANN 8.0.0 之后 op_state 属性从 op_summary 文件中获取
-            if hasattr(profiling_database, "op_summary"):
-                static_shape_operators = profiling_database.op_summary.get_static_shape_operators()
-                if len(static_shape_operators) == 0:
-                    OperatorChecker.IS_ALL_OPERATOR_DYNAMIC_SHAPE = True
-                    return True
-            else:
-                logger.warning(
-                        "Skip dynamic shape checker because of not containing op_summary.csv file in current filefloder."
-                    )
-            
-        return False
+        return self.is_dynamic_shape(profiling_database)
 
     def make_record(self, profiling_database) -> OptimizeRecord:
         """
diff --git a/profiler/advisor/analyzer/computation/operator_checker.py b/profiler/advisor/analyzer/computation/operator_checker.py
index 6bb837004b4282e406d8e1c7b3c5c2a135b9be0c..0f47650943a7355b494bd766214d10526c46c0fa 100644
--- a/profiler/advisor/analyzer/computation/operator_checker.py
+++ b/profiler/advisor/analyzer/computation/operator_checker.py
@@ -15,8 +15,7 @@ logger = logging.getLogger()
 
 
 class OperatorChecker(VersionControl):
-    _SUPPORT_VERSIONS = [constant.CANN_VERSION_C30, constant.CANN_VERSION_C13, constant.CANN_VERSION_C15, constant.CANN_VERSION_C17]
-    IS_ALL_OPERATOR_DYNAMIC_SHAPE = False
+    _SUPPORT_VERSIONS = constant.SUPPORTED_CANN_VERSION
     _MAX_TUNE_OP_NUM = constant.OPERATOR_OUT_TOPK
     _MIN_TASK_DURATION = 0
     _MIN_TASK_DURATION_RATIO = 1.0
@@ -115,10 +114,33 @@ class OperatorChecker(VersionControl):
         return description
 
     def pre_check(self, profiling_data) -> bool:
-        self.format_suggestion_content(profiling_data)
-        return not (OperatorChecker.IS_ALL_OPERATOR_DYNAMIC_SHAPE and (
-                OperatorChecker.PyTorch_OPERATOR_TUNE_SUGGESTION or OperatorChecker.MSLite_OPERATOR_TUNE_SUGGESTION
-        ) in self._SUGGESTION)
+        return True
+
+    def is_dynamic_shape(self, profiling_database: ProfilingDataset) -> bool:
+        less_than_cann800_list = [constant.CANN_VERSION_C30, constant.CANN_VERSION_C13, constant.CANN_VERSION_C15]
+        # CANN 8.0.0 之前从 ge_info 中获取 op_state 属性，进行动态 shape 逻辑判断
+        if self.cann_version in less_than_cann800_list:
+            if hasattr(profiling_database, "ge_info"):
+                ge_info = profiling_database.ge_info
+                static_shape_operators = ge_info.get_static_shape_operators()
+                if len(static_shape_operators) == 0:
+                    return True
+            else:
+                logger.warning(
+                    "Skip dynamic shape check because of not containing ge_info.db file in host filefloder.\n"
+                    "To enable dynamic shape check, please try to set data_simplification=False in experimental_config.\n"
+                    "More details please refer to link : %s", constant.ASCEND_PROFILER_URL)
+        else:
+            # CANN 8.0.0 之后 op_state 属性从 op_summary 文件中获取
+            if hasattr(profiling_database, "op_summary"):
+                static_shape_operators = profiling_database.op_summary.get_static_shape_operators()
+                if len(static_shape_operators) == 0:
+                    return True
+            else:
+                logger.warning(
+                        "Skip dynamic shape check because of not containing op_summary.csv file in current filefloder."
+                    )
+        return False
 
     def format_operator_result(self, record, limit):
         """
@@ -279,4 +301,7 @@ class OperatorChecker(VersionControl):
         return details
 
     def format_suggestion_content(self, profiling_data: ProfilingDataset) -> None:
-        return
+        if profiling_data.PROF_TYPE == constant.ASCEND_PYTORCH_PROFILER:
+            self._SUGGESTION.append(self.PyTorch_OPERATOR_TUNE_SUGGESTION)
+        elif profiling_data.PROF_TYPE == constant.MSLITE:
+            self._SUGGESTION.append(self.MSLite_OPERATOR_TUNE_SUGGESTION)
diff --git a/profiler/advisor/analyzer/computation/profiling_analyzer.py b/profiler/advisor/analyzer/computation/profiling_analyzer.py
index 98d3c5c49b74362137126ec1276c3684284662f0..8682617700702055628a31982b0eafab9feb336d 100644
--- a/profiler/advisor/analyzer/computation/profiling_analyzer.py
+++ b/profiler/advisor/analyzer/computation/profiling_analyzer.py
@@ -22,45 +22,40 @@ class ProfilingAnalyzer(BaseAnalyzer, ABC):
     dataset_cls_list = [ProfilingDataset]
 
     def __init__(self, collection_path, **kwargs) -> None:
-        cann_version = kwargs.get("cann_version", constant.DEFAULT_CANN_VERSION)
-        torch_version = kwargs.get("torch_version", constant.DEFAULT_TORCH_VERSION)
-        super().__init__(collection_path, cann_version=cann_version, torch_version=torch_version, **kwargs)
-        self.checker_list = [checker(cann_version) for checker in get_supported_subclass(OperatorChecker, cann_version)]
-        # 动态 shape checker 放到首位，因为动态 shape 情形下AOE算子调优现在不支持，AOE 算子调优 checker 可以跳过
-        index = next((i for i, item in enumerate(self.checker_list) if isinstance(item, DynamicShapeChecker)), None)
-        self.checker_list.insert(0, self.checker_list.pop(index))
+        super().__init__(collection_path, **kwargs)
+        self.checker = OperatorChecker(self.cann_version)
         self.html_render = HTMLRender()
         self.result = OptimizeResult()
 
     @BaseAnalyzer.check_data((ProfilingDataset.get_key(),))
-    def optimize(self) -> OptimizeResult:
+    def optimize(self, **kwargs) -> OptimizeResult:
         """
         optimize operator
         :param data: input datasets
         :return: result
         """
         profiling_data = self.get_first_data_by_key(self.dataset_list, ProfilingDataset.get_key())
-        for checker in self.checker_list:
-            if not checker.pre_check(profiling_data):
-                continue
-            if checker.check(profiling_data):
-                # add record
-                record = checker.make_record(profiling_data)
-                checker.make_render(self.html_render, record)
-                self.result.add(record)
-                # add details
-                details = checker.get_details()
-                if details:
-                    for i, detail in enumerate(details):
-                        if i == 0:
-                            # the first row is header
-                            self.result.add_detail(checker.get_name(), headers=detail)
-                        else:
-                            self.result.add_detail(checker.get_name(), detail=detail)
-                # add tune op list
-                tune_op_list = checker.get_tune_op_list()
-                if tune_op_list:
-                    self.result.add_tune_op_list(tune_op_list)
+        checker = self.checker
+        if not checker.pre_check(profiling_data):
+            return self.result
+        if checker.check(profiling_data):
+            # add record
+            record = checker.make_record(profiling_data)
+            checker.make_render(self.html_render, record)
+            self.result.add(record)
+            # add details
+            details = checker.get_details()
+            if details:
+                for i, detail in enumerate(details):
+                    if i == 0:
+                        # the first row is header
+                        self.result.add_detail(checker.get_name(), headers=detail)
+                    else:
+                        self.result.add_detail(checker.get_name(), detail=detail)
+            # add tune op list
+            tune_op_list = checker.get_tune_op_list()
+            if tune_op_list:
+                self.result.add_tune_op_list(tune_op_list)
 
         return self.result
 
@@ -69,3 +64,26 @@ class ProfilingAnalyzer(BaseAnalyzer, ABC):
 
     def make_render(self):
         pass
+
+
+class DynamicShapeAnalyzer(ProfilingAnalyzer):
+    def __init__(self, collection_path, **kwargs) -> None:
+        super().__init__(collection_path, **kwargs)
+        self.checker = DynamicShapeChecker(self.cann_version)
+
+
+class BlockDimAnalyzer(ProfilingAnalyzer):
+    def __init__(self, collection_path, **kwargs) -> None:
+        super().__init__(collection_path, **kwargs)
+        self.checker = BlockDimChecker(self.cann_version)
+        
+
+class OperatorBoundAnalyzer(ProfilingAnalyzer):
+    def __init__(self, collection_path, **kwargs) -> None:
+        super().__init__(collection_path, **kwargs)
+        self.checker = OperatorBoundChecker(self.cann_version)
+
+class AicpuAnalyzer(ProfilingAnalyzer):
+    def __init__(self, collection_path, **kwargs) -> None:
+        super().__init__(collection_path, **kwargs)
+        self.checker = AicpuChecker(self.cann_version)
\ No newline at end of file
diff --git a/profiler/advisor/analyzer/graph_fusion/graph_fusion_analyzer.py b/profiler/advisor/analyzer/graph_fusion/graph_fusion_analyzer.py
index 713e1184299944ce506afbd48c4c2f1ec3f7d6e4..326be83b8d49088b1563ccd8c08b68a4aa3001ef 100644
--- a/profiler/advisor/analyzer/graph_fusion/graph_fusion_analyzer.py
+++ b/profiler/advisor/analyzer/graph_fusion/graph_fusion_analyzer.py
@@ -22,7 +22,7 @@ class FusionOPAnalyzer(BaseAnalyzer):
         self.html_render = HTMLRender()
     
     @BaseAnalyzer.check_data((GraphDataset.get_key(),))
-    def optimize(self):
+    def optimize(self, **kwargs):
         """
         :return: result
         """
diff --git a/profiler/advisor/analyzer/overall/overall_analyzer.py b/profiler/advisor/analyzer/overall/overall_analyzer.py
index 7e5102bcfb6937691cfa4b2a962f7aa69c18d35f..916a396b3d096dc788954cbc8e8ba9755cd15f4e 100644
--- a/profiler/advisor/analyzer/overall/overall_analyzer.py
+++ b/profiler/advisor/analyzer/overall/overall_analyzer.py
@@ -18,7 +18,7 @@ class OverallSummaryAnalyzer(BaseAnalyzer):
         self.html_render = HTMLRender()
         self.result = OptimizeResult()
 
-    def optimize(self):
+    def optimize(self, **kwargs):
         compare_result = ComparisonInterface(self.benchmark_profiling_path, self.profiling_path).compare(
             Constant.OVERALL_COMPARE)
 
diff --git a/profiler/advisor/analyzer/overall/overall_summary_analyzer.py b/profiler/advisor/analyzer/overall/overall_summary_analyzer.py
index e810994cd4d4bea76bc506b2121016009f380ecc..c74ae0510331fb9ba8a1794bd724710ba19cfabf 100644
--- a/profiler/advisor/analyzer/overall/overall_summary_analyzer.py
+++ b/profiler/advisor/analyzer/overall/overall_summary_analyzer.py
@@ -53,10 +53,9 @@ class OverallSummaryAnalyzer(BaseAnalyzer):
         "Free Time": ['SDMA Time(Num)']
     }
 
-    def __init__(self, collection_path: str, n_processes: int = 1, cann_version=const.DEFAULT_CANN_VERSION,
-                 torch_version=const.DEFAULT_TORCH_VERSION, **kwargs):
+    def __init__(self, collection_path: str, n_processes: int = 1, **kwargs):
         profile_path = get_profile_path(collection_path)
-        super().__init__(profile_path, n_processes, cann_version, torch_version, **kwargs)
+        super().__init__(profile_path, n_processes, **kwargs)
         self.base_collection_path = kwargs.get("base_collection_path", "")
         self._has_base_collection = False
         self._is_minimal_profiling = False
@@ -179,7 +178,7 @@ class OverallSummaryAnalyzer(BaseAnalyzer):
         self.cur_bottleneck["overall_data"] = overall_bottleneck
         if comparison_bottleneck:
             self.cur_bottleneck["comparison_result"] = comparison_bottleneck
-    def optimize(self):
+    def optimize(self, **kwargs):
         if self.path_check():
             self.process()
         self.identify_bottleneck()
diff --git a/profiler/advisor/analyzer/schedule/fusion_ops/fusion_ops_analyzer.py b/profiler/advisor/analyzer/schedule/fusion_ops/fusion_ops_analyzer.py
index 01613dbe328d513048eb2e1539ac4a19f0c5d587..c1eb24b8e1e11ac167a7eb9333867167a57dd524 100644
--- a/profiler/advisor/analyzer/schedule/fusion_ops/fusion_ops_analyzer.py
+++ b/profiler/advisor/analyzer/schedule/fusion_ops/fusion_ops_analyzer.py
@@ -19,9 +19,8 @@ logger = logging.getLogger()
 class TimelineFusionOpsAnalyzer(BaseAnalyzer):
     dataset_cls_list = [TimelineEventDataset]
 
-    def __init__(self, collection_path, n_processes: int = 1, cann_version=const.DEFAULT_CANN_VERSION,
-                 torch_version=const.DEFAULT_TORCH_VERSION, **kwargs):
-        super().__init__(collection_path, n_processes, cann_version, torch_version, **kwargs)
+    def __init__(self, collection_path, n_processes: int = 1, **kwargs):
+        super().__init__(collection_path, n_processes, **kwargs)
         self._matched_op_index = {} if self.n_processes <= 1 else multiprocessing.Manager().dict()
         self.matched_op_stacks = {}
         self.empty_stacks = True
diff --git a/profiler/advisor/common/analyzer_scopes.py b/profiler/advisor/common/analyzer_scopes.py
index 44f09d0a58b8ec5327846572af65f7161ee93e43..bfe49873a99ad1136b2c385457fb2644df992080 100644
--- a/profiler/advisor/common/analyzer_scopes.py
+++ b/profiler/advisor/common/analyzer_scopes.py
@@ -7,4 +7,7 @@ class SupportedScopes:
     SLOW_RANK = "slow_rank"
     SLOW_LINK = "slow_link"
     OVER_ALL = "over_all"
-    PROFILING_OPERATOR_ANALYSIS = "profiling_operator_analysis"
+    DYNAMIC_SHAPE_ANALYSIS = "dynamic_shape_analysis"
+    AICPU_ANALYSIS = "aicpu_analysis"
+    BLOCK_DIM_ANALYSIS = "block_dim_analysis"
+    OPERATOR_NO_BOUND_ANALYSIS = "operator_no_bound_analysis"
diff --git a/profiler/advisor/computation_analysis.ipynb b/profiler/advisor/computation_analysis.ipynb
new file mode 100644
index 0000000000000000000000000000000000000000..bb9cb1112540a188a62c52a5a83a7c1d210b7571
--- /dev/null
+++ b/profiler/advisor/computation_analysis.ipynb
@@ -0,0 +1,770 @@
+{
+ "cells": [
+  {
+   "cell_type": "code",
+   "execution_count": 2,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from prettytable import PrettyTable, ALL\n",
+    "from textwrap import fill\n",
+    "from profiler.advisor.interface.interface import Interface"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 3,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# 配置profiling采集出来的数据,需要指定到的profiling目录是同一个工具采集的,并且需要采集l0级别以上\n",
+    "profiling_path = r\"YOUR PROFILING PATH\"\n",
+    "interface = Interface(profiling_path=profiling_path)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### Block Dim问题识别\n",
+    "\n",
+    "Block Dim问题主要为识别相关core算子AI core核未打满或者Vector 核未打满问题,主要调优手段为AOE调优,由于AOE调优依赖静态shape,所以当所有算子都为动态shape时,将不会检测相关Block Dim问题.\n",
+    "\n",
+    "下列代码为样例,主要展示如何检测Block Dim类型问题,并获取相关问题检测结果:\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 4,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# 查询computation相关是否存在block dim问题\n",
+    "# 如果profiling数据采集自非8.0.0的CANN版本,需要在训练/推理环境中执行: 'cat CANN安装目录/ascend-toolkit/latest/aarch64-linux/ascend_toolkit_install.info'命令查看version\n",
+    "block_dim_result = interface.get_result(\"computation\", \"block_dim_analysis\", cann_version=\"7.0.RC1\")"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 5,
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/html": [
+       "<table>\n",
+       "    <thead>\n",
+       "        <tr>\n",
+       "            <th>problem</th>\n",
+       "            <th>description</th>\n",
+       "            <th>suggestion</th>\n",
+       "            <th>problem count</th>\n",
+       "            <th>total_time(us)</th>\n",
+       "            <th>time ratio</th>\n",
+       "            <th>income(us)</th>\n",
+       "            <th>income ratio</th>\n",
+       "        </tr>\n",
+       "    </thead>\n",
+       "    <tbody>\n",
+       "        <tr>\n",
+       "            <td>block dim</td>\n",
+       "            <td>some operator does not make full use of 25 ai core or 50 ai vector core;<br> Top-10 operator of task duration are as follows:<br>Square, MatMulV2, BatchMatMul, <br>SoftmaxV2, Mul, Transpose, <br>Assign, GatherV2, Sigmoid, <br>Cast</td>\n",
+       "            <td>1. Optimize operator by AOE, such as:<br>&#x27;aoe --job_type=2 --model_path=$user_dump_path --tune_ops_file=c:\\personalC\\code\\att\\profiler\\advisor\\operator_tuning_file_20240607114012.cfg&#x27;<br></td>\n",
+       "            <td>101</td>\n",
+       "            <td>814.0199999999999</td>\n",
+       "            <td>1.0</td>\n",
+       "            <td></td>\n",
+       "            <td></td>\n",
+       "        </tr>\n",
+       "    </tbody>\n",
+       "</table>"
+      ],
+      "text/plain": [
+       "+-----------+--------------------------------------------------------------------------+------------------------------------------------------------------------------------------------------------------------------------------------+---------------+-------------------+------------+------------+--------------+\n",
+       "| problem   | description                                                              | suggestion                                                                                                                                     | problem count | total_time(us)    | time ratio | income(us) | income ratio |\n",
+       "+-----------+--------------------------------------------------------------------------+------------------------------------------------------------------------------------------------------------------------------------------------+---------------+-------------------+------------+------------+--------------+\n",
+       "| block dim | some operator does not make full use of 25 ai core or 50 ai vector core; | 1. Optimize operator by AOE, such as:                                                                                                          | 101           | 814.0199999999999 | 1.0        |            |              |\n",
+       "|           |  Top-10 operator of task duration are as follows:                        | 'aoe --job_type=2 --model_path=$user_dump_path --tune_ops_file=c:\\personalC\\code\\att\\profiler\\advisor\\operator_tuning_file_20240607114012.cfg' |               |                   |            |            |              |\n",
+       "|           | Square, MatMulV2, BatchMatMul,                                           |                                                                                                                                                |               |                   |            |            |              |\n",
+       "|           | SoftmaxV2, Mul, Transpose,                                               |                                                                                                                                                |               |                   |            |            |              |\n",
+       "|           | Assign, GatherV2, Sigmoid,                                               |                                                                                                                                                |               |                   |            |            |              |\n",
+       "|           | Cast                                                                     |                                                                                                                                                |               |                   |            |            |              |\n",
+       "+-----------+--------------------------------------------------------------------------+------------------------------------------------------------------------------------------------------------------------------------------------+---------------+-------------------+------------+------------+--------------+"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    }
+   ],
+   "source": [
+    "problems = block_dim_result.get(\"problems\")\n",
+    "if problems: # 如果存在相关问题则获取相关问题检测描述及建议\n",
+    "    problem_table = PrettyTable(problems.get(\"headers\"))\n",
+    "    for row in problems.get(\"data\"):\n",
+    "        problem_table.add_row(row)\n",
+    "        \n",
+    "    problem_table.align = \"l\"\n",
+    "    display(problem_table)\n",
+    "else:\n",
+    "    print(\"There is no suggestion related to block dim.\")"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 6,
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/html": [
+       "<table>\n",
+       "    <thead>\n",
+       "        <tr>\n",
+       "            <th>op_name</th>\n",
+       "            <th>op_type</th>\n",
+       "            <th>task_type</th>\n",
+       "            <th>task_duration</th>\n",
+       "            <th>income</th>\n",
+       "            <th>block_dim</th>\n",
+       "            <th>mix_block_dim</th>\n",
+       "            <th>input_shapes</th>\n",
+       "            <th>input_data_types</th>\n",
+       "            <th>input_formats</th>\n",
+       "            <th>output_shapes</th>\n",
+       "            <th>output_data_types</th>\n",
+       "            <th>output_formats</th>\n",
+       "        </tr>\n",
+       "    </thead>\n",
+       "    <tbody>\n",
+       "        <tr>\n",
+       "            <td>Default/model-LlamaModel/layers-<br>CellList/0-LLamaDecodeLayer/attention_norm-<br>LlamaRMSNorm/Square-op34Default/model-<br>LlamaModel/layers-<br>CellList/0-LLamaDecodeLayer/attention_norm-<br>LlamaRMSNorm/ReduceMean-op35</td>\n",
+       "            <td>Square</td>\n",
+       "            <td>AI_VECTOR_CORE</td>\n",
+       "            <td>42.76</td>\n",
+       "            <td>0</td>\n",
+       "            <td>16</td>\n",
+       "            <td>0</td>\n",
+       "            <td>&quot;128,128&quot;</td>\n",
+       "            <td>FLOAT</td>\n",
+       "            <td>NCHW</td>\n",
+       "            <td>&quot;128,1&quot;</td>\n",
+       "            <td>FLOAT</td>\n",
+       "            <td>NCHW</td>\n",
+       "        </tr>\n",
+       "        <tr>\n",
+       "            <td>Default/model-LlamaModel/layers-<br>CellList/0-LLamaDecodeLayer/ffn_norm-<br>LlamaRMSNorm/Square-op77Default/model-<br>LlamaModel/layers-<br>CellList/0-LLamaDecodeLayer/ffn_norm-<br>LlamaRMSNorm/ReduceMean-op78</td>\n",
+       "            <td>Square</td>\n",
+       "            <td>AI_VECTOR_CORE</td>\n",
+       "            <td>42.24</td>\n",
+       "            <td>0</td>\n",
+       "            <td>16</td>\n",
+       "            <td>0</td>\n",
+       "            <td>&quot;128,128&quot;</td>\n",
+       "            <td>FLOAT</td>\n",
+       "            <td>NCHW</td>\n",
+       "            <td>&quot;128,1&quot;</td>\n",
+       "            <td>FLOAT</td>\n",
+       "            <td>NCHW</td>\n",
+       "        </tr>\n",
+       "        <tr>\n",
+       "            <td>Default/lm_head-Linear/MatMul-op213</td>\n",
+       "            <td>MatMulV2</td>\n",
+       "            <td>AI_CORE</td>\n",
+       "            <td>39.02</td>\n",
+       "            <td>0</td>\n",
+       "            <td>20</td>\n",
+       "            <td>0</td>\n",
+       "            <td>&quot;128,128;128,32000&quot;</td>\n",
+       "            <td>FLOAT16;FLOAT16</td>\n",
+       "            <td>FORMAT_ND;FORMAT_ND</td>\n",
+       "            <td>&quot;128,32000&quot;</td>\n",
+       "            <td>FLOAT</td>\n",
+       "            <td>FORMAT_ND</td>\n",
+       "        </tr>\n",
+       "    </tbody>\n",
+       "</table>"
+      ],
+      "text/plain": [
+       "+---------------------------------------------+----------+----------------+---------------+--------+-----------+---------------+---------------------+------------------+---------------------+---------------+-------------------+----------------+\n",
+       "|                   op_name                   | op_type  |   task_type    | task_duration | income | block_dim | mix_block_dim |     input_shapes    | input_data_types |    input_formats    | output_shapes | output_data_types | output_formats |\n",
+       "+---------------------------------------------+----------+----------------+---------------+--------+-----------+---------------+---------------------+------------------+---------------------+---------------+-------------------+----------------+\n",
+       "|       Default/model-LlamaModel/layers-      |  Square  | AI_VECTOR_CORE |     42.76     |   0    |     16    |       0       |      \"128,128\"      |      FLOAT       |         NCHW        |    \"128,1\"    |       FLOAT       |      NCHW      |\n",
+       "| CellList/0-LLamaDecodeLayer/attention_norm- |          |                |               |        |           |               |                     |                  |                     |               |                   |                |\n",
+       "|    LlamaRMSNorm/Square-op34Default/model-   |          |                |               |        |           |               |                     |                  |                     |               |                   |                |\n",
+       "|              LlamaModel/layers-             |          |                |               |        |           |               |                     |                  |                     |               |                   |                |\n",
+       "| CellList/0-LLamaDecodeLayer/attention_norm- |          |                |               |        |           |               |                     |                  |                     |               |                   |                |\n",
+       "|         LlamaRMSNorm/ReduceMean-op35        |          |                |               |        |           |               |                     |                  |                     |               |                   |                |\n",
+       "+---------------------------------------------+----------+----------------+---------------+--------+-----------+---------------+---------------------+------------------+---------------------+---------------+-------------------+----------------+\n",
+       "|       Default/model-LlamaModel/layers-      |  Square  | AI_VECTOR_CORE |     42.24     |   0    |     16    |       0       |      \"128,128\"      |      FLOAT       |         NCHW        |    \"128,1\"    |       FLOAT       |      NCHW      |\n",
+       "|    CellList/0-LLamaDecodeLayer/ffn_norm-    |          |                |               |        |           |               |                     |                  |                     |               |                   |                |\n",
+       "|    LlamaRMSNorm/Square-op77Default/model-   |          |                |               |        |           |               |                     |                  |                     |               |                   |                |\n",
+       "|              LlamaModel/layers-             |          |                |               |        |           |               |                     |                  |                     |               |                   |                |\n",
+       "|    CellList/0-LLamaDecodeLayer/ffn_norm-    |          |                |               |        |           |               |                     |                  |                     |               |                   |                |\n",
+       "|         LlamaRMSNorm/ReduceMean-op78        |          |                |               |        |           |               |                     |                  |                     |               |                   |                |\n",
+       "+---------------------------------------------+----------+----------------+---------------+--------+-----------+---------------+---------------------+------------------+---------------------+---------------+-------------------+----------------+\n",
+       "|     Default/lm_head-Linear/MatMul-op213     | MatMulV2 |    AI_CORE     |     39.02     |   0    |     20    |       0       | \"128,128;128,32000\" | FLOAT16;FLOAT16  | FORMAT_ND;FORMAT_ND |  \"128,32000\"  |       FLOAT       |   FORMAT_ND    |\n",
+       "+---------------------------------------------+----------+----------------+---------------+--------+-----------+---------------+---------------------+------------------+---------------------+---------------+-------------------+----------------+"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    }
+   ],
+   "source": [
+    "if problems: # 如果存在相关问题则获取相关问题检测细节\n",
+    "    block_dim = block_dim_result.get(\"block dim\")\n",
+    "    block_dim_table = PrettyTable(block_dim.get(\"headers\"))\n",
+    "    for row in block_dim.get(\"data\"):\n",
+    "        for i in range(len(row)):\n",
+    "            row[i] = fill(str(row[i]), width=50)\n",
+    "        block_dim_table.add_row(row)\n",
+    "\n",
+    "    block_dim_table.hrules = ALL\n",
+    "    display(block_dim_table[:3])"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### Operator No Bound问题识别\n",
+    "Operator No Bound问题主要为识别相关算子无mte, cube, vector, scalar相关bound问题,主要调优手段为AOE调优,由于AOE调优依赖静态shape,所以当所有算子都为动态shape时,将不会检测相关Operator No Bound问题.\n",
+    "\n",
+    "下列代码为样例,主要展示如何检测Operator No Bound类型问题,并获取相关问题检测结果:"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 7,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from prettytable import PrettyTable, ALL\n",
+    "from textwrap import fill\n",
+    "from profiler.advisor.interface.interface import Interface\n",
+    "\n",
+    "\n",
+    "# 配置profiling采集出来的数据,需要指定到的profiling目录是同一个工具采集的,并且需要采集l0级别以上\n",
+    "profiling_path = r\"YOUR PROFILING PATH\"\n",
+    "interface = Interface(profiling_path=profiling_path)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 8,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# 查询computation相关是否存在operator no bound问题\n",
+    "# 如果profiling数据采集自非8.0.0的CANN版本,需要在训练/推理环境中执行: 'cat CANN安装目录/ascend-toolkit/latest/aarch64-linux/ascend_toolkit_install.info'命令查看version\n",
+    "operator_no_bound_result = interface.get_result(\"computation\", \"operator_no_bound_analysis\", cann_version=\"7.0.RC1\")"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 9,
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/html": [
+       "<table>\n",
+       "    <thead>\n",
+       "        <tr>\n",
+       "            <th>problem</th>\n",
+       "            <th>description</th>\n",
+       "            <th>suggestion</th>\n",
+       "            <th>problem count</th>\n",
+       "            <th>total_time(us)</th>\n",
+       "            <th>time ratio</th>\n",
+       "            <th>income(us)</th>\n",
+       "            <th>income ratio</th>\n",
+       "        </tr>\n",
+       "    </thead>\n",
+       "    <tbody>\n",
+       "        <tr>\n",
+       "            <td>block dim</td>\n",
+       "            <td>some operator does not make full use of 25 ai core or 50 ai vector core;  Top-10 operator of task duration are as<br>follows: Square, MatMulV2, BatchMatMul,  SoftmaxV2, Mul, Transpose,  Assign, GatherV2, Sigmoid,  Cast</td>\n",
+       "            <td>1. Optimize operator by AOE, such as: &#x27;aoe --job_type=2 --model_path=$user_dump_path<br>--tune_ops_file=c:\\personalC\\code\\att\\profiler\\advisor\\operator_tuning_file_20240607114012.cfg&#x27;</td>\n",
+       "            <td>101</td>\n",
+       "            <td>814.0199999999999</td>\n",
+       "            <td>1.0</td>\n",
+       "            <td></td>\n",
+       "            <td></td>\n",
+       "        </tr>\n",
+       "        <tr>\n",
+       "            <td>operator no bound</td>\n",
+       "            <td>There is no mte, cube, vector, scalar ratio is more than 80.00%; Top task duration operators need to be tuned are as<br>follows:  Square, MatMulV2, BatchMatMul,  SoftmaxV2, Mul, Transpose,  Assign, GatherV2, Sigmoid,  Cast</td>\n",
+       "            <td>1. Optimize operator by AOE, such as: &#x27;aoe --job_type=2 --model_path=$user_dump_path<br>--tune_ops_file=c:\\personalC\\code\\att\\profiler\\advisor\\operator_tuning_file_20240607114012.cfg&#x27;</td>\n",
+       "            <td>95</td>\n",
+       "            <td>814.0199999999999</td>\n",
+       "            <td>0.7985</td>\n",
+       "            <td></td>\n",
+       "            <td></td>\n",
+       "        </tr>\n",
+       "    </tbody>\n",
+       "</table>"
+      ],
+      "text/plain": [
+       "+-------------------+----------------------------------------------------------------------------------------------------------------------+-------------------------------------------------------------------------------------------------+---------------+-------------------+------------+------------+--------------+\n",
+       "| problem           | description                                                                                                          | suggestion                                                                                      | problem count | total_time(us)    | time ratio | income(us) | income ratio |\n",
+       "+-------------------+----------------------------------------------------------------------------------------------------------------------+-------------------------------------------------------------------------------------------------+---------------+-------------------+------------+------------+--------------+\n",
+       "| block dim         | some operator does not make full use of 25 ai core or 50 ai vector core;  Top-10 operator of task duration are as    | 1. Optimize operator by AOE, such as: 'aoe --job_type=2 --model_path=$user_dump_path            | 101           | 814.0199999999999 | 1.0        |            |              |\n",
+       "|                   | follows: Square, MatMulV2, BatchMatMul,  SoftmaxV2, Mul, Transpose,  Assign, GatherV2, Sigmoid,  Cast                | --tune_ops_file=c:\\personalC\\code\\att\\profiler\\advisor\\operator_tuning_file_20240607114012.cfg' |               |                   |            |            |              |\n",
+       "| operator no bound | There is no mte, cube, vector, scalar ratio is more than 80.00%; Top task duration operators need to be tuned are as | 1. Optimize operator by AOE, such as: 'aoe --job_type=2 --model_path=$user_dump_path            | 95            | 814.0199999999999 | 0.7985     |            |              |\n",
+       "|                   | follows:  Square, MatMulV2, BatchMatMul,  SoftmaxV2, Mul, Transpose,  Assign, GatherV2, Sigmoid,  Cast               | --tune_ops_file=c:\\personalC\\code\\att\\profiler\\advisor\\operator_tuning_file_20240607114012.cfg' |               |                   |            |            |              |\n",
+       "+-------------------+----------------------------------------------------------------------------------------------------------------------+-------------------------------------------------------------------------------------------------+---------------+-------------------+------------+------------+--------------+"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    }
+   ],
+   "source": [
+    "problems = operator_no_bound_result.get(\"problems\")\n",
+    "problem_table = PrettyTable(problems.get(\"headers\"))\n",
+    "if problems: # 如果存在相关问题则获取相关问题检测描述及建议\n",
+    "    for row in problems.get(\"data\"):\n",
+    "        for i in range(len(row)):\n",
+    "            row[i] = fill(str(row[i]), width=120)\n",
+    "        problem_table.add_row(row)\n",
+    "\n",
+    "    problem_table.align = \"l\"\n",
+    "    display(problem_table)\n",
+    "else:\n",
+    "    print(\"There is no suggestion related to operator no bound.\")"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 10,
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/html": [
+       "<table>\n",
+       "    <thead>\n",
+       "        <tr>\n",
+       "            <th>op_name</th>\n",
+       "            <th>op_type</th>\n",
+       "            <th>task_type</th>\n",
+       "            <th>task_duration</th>\n",
+       "            <th>vec_ratio</th>\n",
+       "            <th>mac_ratio</th>\n",
+       "            <th>scalar_ratio</th>\n",
+       "            <th>mte1_ratio</th>\n",
+       "            <th>mte2_ratio</th>\n",
+       "            <th>mte3_ratio</th>\n",
+       "            <th>block_dim</th>\n",
+       "            <th>input_shapes</th>\n",
+       "            <th>input_data_types</th>\n",
+       "            <th>input_formats</th>\n",
+       "            <th>output_shapes</th>\n",
+       "            <th>output_data_types</th>\n",
+       "            <th>output_formats</th>\n",
+       "        </tr>\n",
+       "    </thead>\n",
+       "    <tbody>\n",
+       "        <tr>\n",
+       "            <td>Default/model-LlamaModel/layers-<br>CellList/0-LLamaDecodeLayer/attention_norm-<br>LlamaRMSNorm/Square-op34Default/model-LlamaModel/layers-<br>CellList/0-LLamaDecodeLayer/attention_norm-<br>LlamaRMSNorm/ReduceMean-op35</td>\n",
+       "            <td>Square</td>\n",
+       "            <td>AI_VECTOR_CORE</td>\n",
+       "            <td>42.76</td>\n",
+       "            <td>0.4654</td>\n",
+       "            <td>0</td>\n",
+       "            <td>0</td>\n",
+       "            <td>0</td>\n",
+       "            <td>0</td>\n",
+       "            <td>0.0056</td>\n",
+       "            <td>16</td>\n",
+       "            <td>&quot;128,128&quot;</td>\n",
+       "            <td>FLOAT</td>\n",
+       "            <td>NCHW</td>\n",
+       "            <td>&quot;128,1&quot;</td>\n",
+       "            <td>FLOAT</td>\n",
+       "            <td>NCHW</td>\n",
+       "        </tr>\n",
+       "        <tr>\n",
+       "            <td>Default/model-LlamaModel/layers-<br>CellList/0-LLamaDecodeLayer/ffn_norm-LlamaRMSNorm/Square-<br>op77Default/model-LlamaModel/layers-<br>CellList/0-LLamaDecodeLayer/ffn_norm-<br>LlamaRMSNorm/ReduceMean-op78</td>\n",
+       "            <td>Square</td>\n",
+       "            <td>AI_VECTOR_CORE</td>\n",
+       "            <td>42.24</td>\n",
+       "            <td>0.466</td>\n",
+       "            <td>0</td>\n",
+       "            <td>0</td>\n",
+       "            <td>0</td>\n",
+       "            <td>0</td>\n",
+       "            <td>0.0062</td>\n",
+       "            <td>16</td>\n",
+       "            <td>&quot;128,128&quot;</td>\n",
+       "            <td>FLOAT</td>\n",
+       "            <td>NCHW</td>\n",
+       "            <td>&quot;128,1&quot;</td>\n",
+       "            <td>FLOAT</td>\n",
+       "            <td>NCHW</td>\n",
+       "        </tr>\n",
+       "        <tr>\n",
+       "            <td>Default/lm_head-Linear/MatMul-op213</td>\n",
+       "            <td>MatMulV2</td>\n",
+       "            <td>AI_CORE</td>\n",
+       "            <td>39.02</td>\n",
+       "            <td>0</td>\n",
+       "            <td>0.1105</td>\n",
+       "            <td>0.0119</td>\n",
+       "            <td>0.0857</td>\n",
+       "            <td>0.4284</td>\n",
+       "            <td>0</td>\n",
+       "            <td>20</td>\n",
+       "            <td>&quot;128,128;128,32000&quot;</td>\n",
+       "            <td>FLOAT16;FLOAT16</td>\n",
+       "            <td>FORMAT_ND;FORMAT_ND</td>\n",
+       "            <td>&quot;128,32000&quot;</td>\n",
+       "            <td>FLOAT</td>\n",
+       "            <td>FORMAT_ND</td>\n",
+       "        </tr>\n",
+       "    </tbody>\n",
+       "</table>"
+      ],
+      "text/plain": [
+       "+-----------------------------------------------------------+----------+----------------+---------------+-----------+-----------+--------------+------------+------------+------------+-----------+---------------------+------------------+---------------------+---------------+-------------------+----------------+\n",
+       "|                          op_name                          | op_type  |   task_type    | task_duration | vec_ratio | mac_ratio | scalar_ratio | mte1_ratio | mte2_ratio | mte3_ratio | block_dim |     input_shapes    | input_data_types |    input_formats    | output_shapes | output_data_types | output_formats |\n",
+       "+-----------------------------------------------------------+----------+----------------+---------------+-----------+-----------+--------------+------------+------------+------------+-----------+---------------------+------------------+---------------------+---------------+-------------------+----------------+\n",
+       "|              Default/model-LlamaModel/layers-             |  Square  | AI_VECTOR_CORE |     42.76     |   0.4654  |     0     |      0       |     0      |     0      |   0.0056   |     16    |      \"128,128\"      |      FLOAT       |         NCHW        |    \"128,1\"    |       FLOAT       |      NCHW      |\n",
+       "|        CellList/0-LLamaDecodeLayer/attention_norm-        |          |                |               |           |           |              |            |            |            |           |                     |                  |                     |               |                   |                |\n",
+       "|  LlamaRMSNorm/Square-op34Default/model-LlamaModel/layers- |          |                |               |           |           |              |            |            |            |           |                     |                  |                     |               |                   |                |\n",
+       "|        CellList/0-LLamaDecodeLayer/attention_norm-        |          |                |               |           |           |              |            |            |            |           |                     |                  |                     |               |                   |                |\n",
+       "|                LlamaRMSNorm/ReduceMean-op35               |          |                |               |           |           |              |            |            |            |           |                     |                  |                     |               |                   |                |\n",
+       "+-----------------------------------------------------------+----------+----------------+---------------+-----------+-----------+--------------+------------+------------+------------+-----------+---------------------+------------------+---------------------+---------------+-------------------+----------------+\n",
+       "|              Default/model-LlamaModel/layers-             |  Square  | AI_VECTOR_CORE |     42.24     |   0.466   |     0     |      0       |     0      |     0      |   0.0062   |     16    |      \"128,128\"      |      FLOAT       |         NCHW        |    \"128,1\"    |       FLOAT       |      NCHW      |\n",
+       "| CellList/0-LLamaDecodeLayer/ffn_norm-LlamaRMSNorm/Square- |          |                |               |           |           |              |            |            |            |           |                     |                  |                     |               |                   |                |\n",
+       "|            op77Default/model-LlamaModel/layers-           |          |                |               |           |           |              |            |            |            |           |                     |                  |                     |               |                   |                |\n",
+       "|           CellList/0-LLamaDecodeLayer/ffn_norm-           |          |                |               |           |           |              |            |            |            |           |                     |                  |                     |               |                   |                |\n",
+       "|                LlamaRMSNorm/ReduceMean-op78               |          |                |               |           |           |              |            |            |            |           |                     |                  |                     |               |                   |                |\n",
+       "+-----------------------------------------------------------+----------+----------------+---------------+-----------+-----------+--------------+------------+------------+------------+-----------+---------------------+------------------+---------------------+---------------+-------------------+----------------+\n",
+       "|            Default/lm_head-Linear/MatMul-op213            | MatMulV2 |    AI_CORE     |     39.02     |     0     |   0.1105  |    0.0119    |   0.0857   |   0.4284   |     0      |     20    | \"128,128;128,32000\" | FLOAT16;FLOAT16  | FORMAT_ND;FORMAT_ND |  \"128,32000\"  |       FLOAT       |   FORMAT_ND    |\n",
+       "+-----------------------------------------------------------+----------+----------------+---------------+-----------+-----------+--------------+------------+------------+------------+-----------+---------------------+------------------+---------------------+---------------+-------------------+----------------+"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    }
+   ],
+   "source": [
+    "if problems: # 如果存在相关问题则获取相关问题检测细节\n",
+    "    operator_no_bound = operator_no_bound_result.get(\"operator no bound\")\n",
+    "    operator_no_bound_table = PrettyTable(operator_no_bound.get(\"headers\"))\n",
+    "    for row in operator_no_bound.get(\"data\"):\n",
+    "        for i in range(len(row)):\n",
+    "            row[i] = fill(str(row[i]), width=60)\n",
+    "        operator_no_bound_table.add_row(row)\n",
+    "    operator_no_bound_table.hrules = ALL\n",
+    "    display(operator_no_bound_table[:3])"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### AICPU问题识别\n",
+    "AICPU问题主要为识别相关算子执行时跑到AICPU上计算,并没有利用到AI CORE的计算能力的场景,主要调优手段为修改相关代码来避免AICPU算子,可参见相关资料,来避免AICPU算子的问题:\n",
+    "https://support.huaweicloud.com/bestpractice-modelarts/modelarts_10_2517.html\n",
+    "\n",
+    "下列代码为样例,主要展示如何检测Dynamic Shape类型问题,并获取相关问题检测结果:"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 11,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from prettytable import PrettyTable, ALL\n",
+    "from textwrap import fill\n",
+    "from profiler.advisor.interface.interface import Interface\n",
+    "\n",
+    "\n",
+    "# 配置profiling采集出来的数据,需要指定到的profiling目录是同一个工具采集的,并且需要采集l0级别以上\n",
+    "profiling_path = r\"YOUR PROFILING PATH\"\n",
+    "interface = Interface(profiling_path=profiling_path)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 12,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "Please ensure only one trace_view.json in C:\\personalC\\profiling_data, there will analyze first timeline profiling data.\n",
+      "                                                                                                    \r"
+     ]
+    }
+   ],
+   "source": [
+    "# 查询computation相关是否存在aicpu问题\n",
+    "# 如果profiling数据采集自非8.0.0的CANN版本,需要在训练/推理环境中执行: 'cat CANN安装目录/ascend-toolkit/latest/aarch64-linux/ascend_toolkit_install.info'命令查看version\n",
+    "aicpu_result = interface.get_result(\"computation\", \"aicpu_analysis\")"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 13,
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/html": [
+       "<table>\n",
+       "    <thead>\n",
+       "        <tr>\n",
+       "            <th>problem</th>\n",
+       "            <th>description</th>\n",
+       "            <th>suggestion</th>\n",
+       "            <th>problem count</th>\n",
+       "            <th>total_time(us)</th>\n",
+       "            <th>time ratio</th>\n",
+       "            <th>income(us)</th>\n",
+       "            <th>income ratio</th>\n",
+       "        </tr>\n",
+       "    </thead>\n",
+       "    <tbody>\n",
+       "        <tr>\n",
+       "            <td>block dim</td>\n",
+       "            <td>some operator does not make full use of 25 ai core or 50 ai vector core;  Top-10 operator of task duration are as<br>follows: Square, MatMulV2, BatchMatMul,  SoftmaxV2, Mul, Transpose,  Assign, GatherV2, Sigmoid,  Cast</td>\n",
+       "            <td>1. Optimize operator by AOE, such as: &#x27;aoe --job_type=2 --model_path=$user_dump_path<br>--tune_ops_file=c:\\personalC\\code\\att\\profiler\\advisor\\operator_tuning_file_20240607114012.cfg&#x27;</td>\n",
+       "            <td>101</td>\n",
+       "            <td>814.0199999999999</td>\n",
+       "            <td>1.0</td>\n",
+       "            <td></td>\n",
+       "            <td></td>\n",
+       "        </tr>\n",
+       "        <tr>\n",
+       "            <td>operator no bound</td>\n",
+       "            <td>There is no mte, cube, vector, scalar ratio is more than 80.00%; Top task duration operators need to be tuned are as<br>follows:  Square, MatMulV2, BatchMatMul,  SoftmaxV2, Mul, Transpose,  Assign, GatherV2, Sigmoid,  Cast</td>\n",
+       "            <td>1. Optimize operator by AOE, such as: &#x27;aoe --job_type=2 --model_path=$user_dump_path<br>--tune_ops_file=c:\\personalC\\code\\att\\profiler\\advisor\\operator_tuning_file_20240607114012.cfg&#x27;</td>\n",
+       "            <td>95</td>\n",
+       "            <td>814.0199999999999</td>\n",
+       "            <td>0.7985</td>\n",
+       "            <td></td>\n",
+       "            <td></td>\n",
+       "        </tr>\n",
+       "        <tr>\n",
+       "            <td>AICPU operator</td>\n",
+       "            <td>Some operators and task duration exceed 20 us, such as :<br>Cast</td>\n",
+       "            <td>1. Modify code to avoid aicpu operator</td>\n",
+       "            <td>39</td>\n",
+       "            <td>686568.860000001</td>\n",
+       "            <td>0.0189</td>\n",
+       "            <td></td>\n",
+       "            <td></td>\n",
+       "        </tr>\n",
+       "    </tbody>\n",
+       "</table>"
+      ],
+      "text/plain": [
+       "+-------------------+----------------------------------------------------------------------------------------------------------------------+-------------------------------------------------------------------------------------------------+---------------+-------------------+------------+------------+--------------+\n",
+       "| problem           | description                                                                                                          | suggestion                                                                                      | problem count | total_time(us)    | time ratio | income(us) | income ratio |\n",
+       "+-------------------+----------------------------------------------------------------------------------------------------------------------+-------------------------------------------------------------------------------------------------+---------------+-------------------+------------+------------+--------------+\n",
+       "| block dim         | some operator does not make full use of 25 ai core or 50 ai vector core;  Top-10 operator of task duration are as    | 1. Optimize operator by AOE, such as: 'aoe --job_type=2 --model_path=$user_dump_path            | 101           | 814.0199999999999 | 1.0        |            |              |\n",
+       "|                   | follows: Square, MatMulV2, BatchMatMul,  SoftmaxV2, Mul, Transpose,  Assign, GatherV2, Sigmoid,  Cast                | --tune_ops_file=c:\\personalC\\code\\att\\profiler\\advisor\\operator_tuning_file_20240607114012.cfg' |               |                   |            |            |              |\n",
+       "| operator no bound | There is no mte, cube, vector, scalar ratio is more than 80.00%; Top task duration operators need to be tuned are as | 1. Optimize operator by AOE, such as: 'aoe --job_type=2 --model_path=$user_dump_path            | 95            | 814.0199999999999 | 0.7985     |            |              |\n",
+       "|                   | follows:  Square, MatMulV2, BatchMatMul,  SoftmaxV2, Mul, Transpose,  Assign, GatherV2, Sigmoid,  Cast               | --tune_ops_file=c:\\personalC\\code\\att\\profiler\\advisor\\operator_tuning_file_20240607114012.cfg' |               |                   |            |            |              |\n",
+       "| AICPU operator    | Some operators and task duration exceed 20 us, such as :                                                             | 1. Modify code to avoid aicpu operator                                                          | 39            | 686568.860000001  | 0.0189     |            |              |\n",
+       "|                   | Cast                                                                                                                 |                                                                                                 |               |                   |            |            |              |\n",
+       "+-------------------+----------------------------------------------------------------------------------------------------------------------+-------------------------------------------------------------------------------------------------+---------------+-------------------+------------+------------+--------------+"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    }
+   ],
+   "source": [
+    "problems = aicpu_result.get(\"problems\")\n",
+    "if problems: # 如果存在相关问题则获取相关问题检测描述及建议\n",
+    "    problem_table = PrettyTable(problems.get(\"headers\"))\n",
+    "    for row in problems.get(\"data\"):\n",
+    "        problem_table.add_row(row)\n",
+    "\n",
+    "    problem_table.align = \"l\"\n",
+    "    display(problem_table)\n",
+    "else:\n",
+    "    print(\"There is no suggestion related to operator no bound.\")"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 14,
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/html": [
+       "<table>\n",
+       "    <thead>\n",
+       "        <tr>\n",
+       "            <th>op_name</th>\n",
+       "            <th>op_type</th>\n",
+       "            <th>task_duration</th>\n",
+       "            <th>input_shapes</th>\n",
+       "            <th>input_data_types</th>\n",
+       "            <th>input_formats</th>\n",
+       "            <th>output_shapes</th>\n",
+       "            <th>output_data_types</th>\n",
+       "            <th>output_formats</th>\n",
+       "            <th>stack_info</th>\n",
+       "        </tr>\n",
+       "    </thead>\n",
+       "    <tbody>\n",
+       "        <tr>\n",
+       "            <td>trans_Cast_5</td>\n",
+       "            <td>Cast</td>\n",
+       "            <td>493.64</td>\n",
+       "            <td>&quot;&quot;</td>\n",
+       "            <td>INT32</td>\n",
+       "            <td>FORMAT_ND</td>\n",
+       "            <td>&quot;&quot;</td>\n",
+       "            <td>UINT64</td>\n",
+       "            <td>FORMAT_ND</td>\n",
+       "            <td>/usr/local/python3.7.5/lib/python3.7/site-<br>packages/torch/nn/functional.py(1279): dropout;<br>/usr/local/python3.7.5/lib/python3.7/site-<br>packages/torch/nn/modules/dropout.py(58): forward;<br>/usr/local/python3.7.5/lib/python3.7/site-<br>packages/torch/nn/modules/module.py(1110):<br>_call_impl;  /profiling_auto_GPT3/megatron/model/l<br>anguage_model.py(236): forward;<br>/usr/local/python3.7.5/lib/python3.7/site-<br>packages/torch/nn/modules/module.py(1110):<br>_call_impl;  /profiling_auto_GPT3/megatron/model/l<br>anguage_model.py(425): forward;<br>/usr/local/python3.7.5/lib/python3.7/site-<br>packages/torch/nn/modules/module.py(1110):<br>_call_impl;  /profiling_auto_GPT3/megatron/model/g<br>pt_model.py(84): forward;<br>/usr/local/python3.7.5/lib/python3.7/site-<br>packages/torch/nn/modules/module.py(1110):<br>_call_impl;  /profiling_auto_GPT3/megatron/model/m<br>odule.py(184): forward;<br>/usr/local/python3.7.5/lib/python3.7/site-<br>packages/torch/nn/modules/module.py(1110):<br>_call_impl;  /profiling_auto_GPT3/megatron/model/d<br>istributed.py(58): forward;<br>/usr/local/python3.7.5/lib/python3.7/site-<br>packages/torch/nn/modules/module.py(1110):<br>_call_impl;  ../../pretrain_gpt.py(88):<br>forward_step;<br>/profiling_auto_GPT3/megatron/schedules.py(118):<br>forward_step;  /home/s30040711/Megatron-LM/megatro<br>n_npu_adaptor/megatron_npu/adaptor_schedules.py(96<br>): forward_backward_no_pipelining;<br>/profiling_auto_GPT3/megatron/training.py(419):<br>train_step;<br>/profiling_auto_GPT3/megatron/training.py(837):<br>train;<br>/profiling_auto_GPT3/megatron/training.py(152):<br>pretrain;  ../../pretrain_gpt.py(122): &lt;module&gt;</td>\n",
+       "        </tr>\n",
+       "        <tr>\n",
+       "            <td>trans_Cast_5</td>\n",
+       "            <td>Cast</td>\n",
+       "            <td>413.4</td>\n",
+       "            <td>&quot;&quot;</td>\n",
+       "            <td>INT32</td>\n",
+       "            <td>FORMAT_ND</td>\n",
+       "            <td>&quot;&quot;</td>\n",
+       "            <td>UINT64</td>\n",
+       "            <td>FORMAT_ND</td>\n",
+       "            <td>/usr/local/python3.7.5/lib/python3.7/site-<br>packages/torch/nn/functional.py(1279): dropout;<br>/usr/local/python3.7.5/lib/python3.7/site-<br>packages/torch/nn/modules/dropout.py(58): forward;<br>/usr/local/python3.7.5/lib/python3.7/site-<br>packages/torch/nn/modules/module.py(1110):<br>_call_impl;  /profiling_auto_GPT3/megatron/model/l<br>anguage_model.py(236): forward;<br>/usr/local/python3.7.5/lib/python3.7/site-<br>packages/torch/nn/modules/module.py(1110):<br>_call_impl;  /profiling_auto_GPT3/megatron/model/l<br>anguage_model.py(425): forward;<br>/usr/local/python3.7.5/lib/python3.7/site-<br>packages/torch/nn/modules/module.py(1110):<br>_call_impl;  /profiling_auto_GPT3/megatron/model/g<br>pt_model.py(84): forward;<br>/usr/local/python3.7.5/lib/python3.7/site-<br>packages/torch/nn/modules/module.py(1110):<br>_call_impl;  /profiling_auto_GPT3/megatron/model/m<br>odule.py(184): forward;<br>/usr/local/python3.7.5/lib/python3.7/site-<br>packages/torch/nn/modules/module.py(1110):<br>_call_impl;  /profiling_auto_GPT3/megatron/model/d<br>istributed.py(58): forward;<br>/usr/local/python3.7.5/lib/python3.7/site-<br>packages/torch/nn/modules/module.py(1110):<br>_call_impl;  ../../pretrain_gpt.py(88):<br>forward_step;<br>/profiling_auto_GPT3/megatron/schedules.py(118):<br>forward_step;  /home/s30040711/Megatron-LM/megatro<br>n_npu_adaptor/megatron_npu/adaptor_schedules.py(10<br>9): forward_backward_no_pipelining;<br>/profiling_auto_GPT3/megatron/training.py(419):<br>train_step;<br>/profiling_auto_GPT3/megatron/training.py(837):<br>train;<br>/profiling_auto_GPT3/megatron/training.py(152):<br>pretrain;  ../../pretrain_gpt.py(122): &lt;module&gt;</td>\n",
+       "        </tr>\n",
+       "    </tbody>\n",
+       "</table>"
+      ],
+      "text/plain": [
+       "+--------------+---------+---------------+--------------+------------------+---------------+---------------+-------------------+----------------+----------------------------------------------------+\n",
+       "|   op_name    | op_type | task_duration | input_shapes | input_data_types | input_formats | output_shapes | output_data_types | output_formats |                     stack_info                     |\n",
+       "+--------------+---------+---------------+--------------+------------------+---------------+---------------+-------------------+----------------+----------------------------------------------------+\n",
+       "| trans_Cast_5 |   Cast  |     493.64    |      \"\"      |      INT32       |   FORMAT_ND   |       \"\"      |       UINT64      |   FORMAT_ND    |     /usr/local/python3.7.5/lib/python3.7/site-     |\n",
+       "|              |         |               |              |                  |               |               |                   |                |  packages/torch/nn/functional.py(1279): dropout;   |\n",
+       "|              |         |               |              |                  |               |               |                   |                |     /usr/local/python3.7.5/lib/python3.7/site-     |\n",
+       "|              |         |               |              |                  |               |               |                   |                | packages/torch/nn/modules/dropout.py(58): forward; |\n",
+       "|              |         |               |              |                  |               |               |                   |                |     /usr/local/python3.7.5/lib/python3.7/site-     |\n",
+       "|              |         |               |              |                  |               |               |                   |                |     packages/torch/nn/modules/module.py(1110):     |\n",
+       "|              |         |               |              |                  |               |               |                   |                | _call_impl;  /profiling_auto_GPT3/megatron/model/l |\n",
+       "|              |         |               |              |                  |               |               |                   |                |          anguage_model.py(236): forward;           |\n",
+       "|              |         |               |              |                  |               |               |                   |                |     /usr/local/python3.7.5/lib/python3.7/site-     |\n",
+       "|              |         |               |              |                  |               |               |                   |                |     packages/torch/nn/modules/module.py(1110):     |\n",
+       "|              |         |               |              |                  |               |               |                   |                | _call_impl;  /profiling_auto_GPT3/megatron/model/l |\n",
+       "|              |         |               |              |                  |               |               |                   |                |          anguage_model.py(425): forward;           |\n",
+       "|              |         |               |              |                  |               |               |                   |                |     /usr/local/python3.7.5/lib/python3.7/site-     |\n",
+       "|              |         |               |              |                  |               |               |                   |                |     packages/torch/nn/modules/module.py(1110):     |\n",
+       "|              |         |               |              |                  |               |               |                   |                | _call_impl;  /profiling_auto_GPT3/megatron/model/g |\n",
+       "|              |         |               |              |                  |               |               |                   |                |             pt_model.py(84): forward;              |\n",
+       "|              |         |               |              |                  |               |               |                   |                |     /usr/local/python3.7.5/lib/python3.7/site-     |\n",
+       "|              |         |               |              |                  |               |               |                   |                |     packages/torch/nn/modules/module.py(1110):     |\n",
+       "|              |         |               |              |                  |               |               |                   |                | _call_impl;  /profiling_auto_GPT3/megatron/model/m |\n",
+       "|              |         |               |              |                  |               |               |                   |                |              odule.py(184): forward;               |\n",
+       "|              |         |               |              |                  |               |               |                   |                |     /usr/local/python3.7.5/lib/python3.7/site-     |\n",
+       "|              |         |               |              |                  |               |               |                   |                |     packages/torch/nn/modules/module.py(1110):     |\n",
+       "|              |         |               |              |                  |               |               |                   |                | _call_impl;  /profiling_auto_GPT3/megatron/model/d |\n",
+       "|              |         |               |              |                  |               |               |                   |                |            istributed.py(58): forward;             |\n",
+       "|              |         |               |              |                  |               |               |                   |                |     /usr/local/python3.7.5/lib/python3.7/site-     |\n",
+       "|              |         |               |              |                  |               |               |                   |                |     packages/torch/nn/modules/module.py(1110):     |\n",
+       "|              |         |               |              |                  |               |               |                   |                |      _call_impl;  ../../pretrain_gpt.py(88):       |\n",
+       "|              |         |               |              |                  |               |               |                   |                |                   forward_step;                    |\n",
+       "|              |         |               |              |                  |               |               |                   |                |  /profiling_auto_GPT3/megatron/schedules.py(118):  |\n",
+       "|              |         |               |              |                  |               |               |                   |                | forward_step;  /home/s30040711/Megatron-LM/megatro |\n",
+       "|              |         |               |              |                  |               |               |                   |                | n_npu_adaptor/megatron_npu/adaptor_schedules.py(96 |\n",
+       "|              |         |               |              |                  |               |               |                   |                |         ): forward_backward_no_pipelining;         |\n",
+       "|              |         |               |              |                  |               |               |                   |                |  /profiling_auto_GPT3/megatron/training.py(419):   |\n",
+       "|              |         |               |              |                  |               |               |                   |                |                    train_step;                     |\n",
+       "|              |         |               |              |                  |               |               |                   |                |  /profiling_auto_GPT3/megatron/training.py(837):   |\n",
+       "|              |         |               |              |                  |               |               |                   |                |                       train;                       |\n",
+       "|              |         |               |              |                  |               |               |                   |                |  /profiling_auto_GPT3/megatron/training.py(152):   |\n",
+       "|              |         |               |              |                  |               |               |                   |                |  pretrain;  ../../pretrain_gpt.py(122): <module>   |\n",
+       "+--------------+---------+---------------+--------------+------------------+---------------+---------------+-------------------+----------------+----------------------------------------------------+\n",
+       "| trans_Cast_5 |   Cast  |     413.4     |      \"\"      |      INT32       |   FORMAT_ND   |       \"\"      |       UINT64      |   FORMAT_ND    |     /usr/local/python3.7.5/lib/python3.7/site-     |\n",
+       "|              |         |               |              |                  |               |               |                   |                |  packages/torch/nn/functional.py(1279): dropout;   |\n",
+       "|              |         |               |              |                  |               |               |                   |                |     /usr/local/python3.7.5/lib/python3.7/site-     |\n",
+       "|              |         |               |              |                  |               |               |                   |                | packages/torch/nn/modules/dropout.py(58): forward; |\n",
+       "|              |         |               |              |                  |               |               |                   |                |     /usr/local/python3.7.5/lib/python3.7/site-     |\n",
+       "|              |         |               |              |                  |               |               |                   |                |     packages/torch/nn/modules/module.py(1110):     |\n",
+       "|              |         |               |              |                  |               |               |                   |                | _call_impl;  /profiling_auto_GPT3/megatron/model/l |\n",
+       "|              |         |               |              |                  |               |               |                   |                |          anguage_model.py(236): forward;           |\n",
+       "|              |         |               |              |                  |               |               |                   |                |     /usr/local/python3.7.5/lib/python3.7/site-     |\n",
+       "|              |         |               |              |                  |               |               |                   |                |     packages/torch/nn/modules/module.py(1110):     |\n",
+       "|              |         |               |              |                  |               |               |                   |                | _call_impl;  /profiling_auto_GPT3/megatron/model/l |\n",
+       "|              |         |               |              |                  |               |               |                   |                |          anguage_model.py(425): forward;           |\n",
+       "|              |         |               |              |                  |               |               |                   |                |     /usr/local/python3.7.5/lib/python3.7/site-     |\n",
+       "|              |         |               |              |                  |               |               |                   |                |     packages/torch/nn/modules/module.py(1110):     |\n",
+       "|              |         |               |              |                  |               |               |                   |                | _call_impl;  /profiling_auto_GPT3/megatron/model/g |\n",
+       "|              |         |               |              |                  |               |               |                   |                |             pt_model.py(84): forward;              |\n",
+       "|              |         |               |              |                  |               |               |                   |                |     /usr/local/python3.7.5/lib/python3.7/site-     |\n",
+       "|              |         |               |              |                  |               |               |                   |                |     packages/torch/nn/modules/module.py(1110):     |\n",
+       "|              |         |               |              |                  |               |               |                   |                | _call_impl;  /profiling_auto_GPT3/megatron/model/m |\n",
+       "|              |         |               |              |                  |               |               |                   |                |              odule.py(184): forward;               |\n",
+       "|              |         |               |              |                  |               |               |                   |                |     /usr/local/python3.7.5/lib/python3.7/site-     |\n",
+       "|              |         |               |              |                  |               |               |                   |                |     packages/torch/nn/modules/module.py(1110):     |\n",
+       "|              |         |               |              |                  |               |               |                   |                | _call_impl;  /profiling_auto_GPT3/megatron/model/d |\n",
+       "|              |         |               |              |                  |               |               |                   |                |            istributed.py(58): forward;             |\n",
+       "|              |         |               |              |                  |               |               |                   |                |     /usr/local/python3.7.5/lib/python3.7/site-     |\n",
+       "|              |         |               |              |                  |               |               |                   |                |     packages/torch/nn/modules/module.py(1110):     |\n",
+       "|              |         |               |              |                  |               |               |                   |                |      _call_impl;  ../../pretrain_gpt.py(88):       |\n",
+       "|              |         |               |              |                  |               |               |                   |                |                   forward_step;                    |\n",
+       "|              |         |               |              |                  |               |               |                   |                |  /profiling_auto_GPT3/megatron/schedules.py(118):  |\n",
+       "|              |         |               |              |                  |               |               |                   |                | forward_step;  /home/s30040711/Megatron-LM/megatro |\n",
+       "|              |         |               |              |                  |               |               |                   |                | n_npu_adaptor/megatron_npu/adaptor_schedules.py(10 |\n",
+       "|              |         |               |              |                  |               |               |                   |                |        9): forward_backward_no_pipelining;         |\n",
+       "|              |         |               |              |                  |               |               |                   |                |  /profiling_auto_GPT3/megatron/training.py(419):   |\n",
+       "|              |         |               |              |                  |               |               |                   |                |                    train_step;                     |\n",
+       "|              |         |               |              |                  |               |               |                   |                |  /profiling_auto_GPT3/megatron/training.py(837):   |\n",
+       "|              |         |               |              |                  |               |               |                   |                |                       train;                       |\n",
+       "|              |         |               |              |                  |               |               |                   |                |  /profiling_auto_GPT3/megatron/training.py(152):   |\n",
+       "|              |         |               |              |                  |               |               |                   |                |  pretrain;  ../../pretrain_gpt.py(122): <module>   |\n",
+       "+--------------+---------+---------------+--------------+------------------+---------------+---------------+-------------------+----------------+----------------------------------------------------+"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    }
+   ],
+   "source": [
+    "if problems: # 如果存在相关问题则获取相关问题检测细节\n",
+    "    aicpu = aicpu_result.get(\"AICPU operator\")\n",
+    "    aicpu_table = PrettyTable(aicpu.get(\"headers\"))\n",
+    "    for row in aicpu.get(\"data\"):\n",
+    "        for i in range(len(row)):\n",
+    "            row[i] = fill(str(row[i]), width=50)\n",
+    "        aicpu_table.add_row(row)\n",
+    "    aicpu_table.hrules = ALL\n",
+    "    display(aicpu_table[:2])"
+   ]
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "base",
+   "language": "python",
+   "name": "python3"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.9.7"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 2
+}
diff --git a/profiler/advisor/dataset/cluster/cluster_dataset.py b/profiler/advisor/dataset/cluster/cluster_dataset.py
index 466349afd7378cb980a37e6a391a486fc0e9bc13..09fda2d4dcf2df2f05abb0007befb5c5c36ef824 100644
--- a/profiler/advisor/dataset/cluster/cluster_dataset.py
+++ b/profiler/advisor/dataset/cluster/cluster_dataset.py
@@ -65,7 +65,7 @@ class ClusterDataset(Dataset):
 class ClusterStepTraceTimeDataSet(ClusterDataset):
     RANK = "rank"
 
-    def __init__(self, collection_path: str, data: dict, kwargs: dict = None):
+    def __init__(self, collection_path: str, data: dict, **kwargs):
         self._step_dict = defaultdict()
         super().__init__(collection_path, data)
 
@@ -107,7 +107,7 @@ class ClusterCommunicationDataSet(ClusterDataset):
     SDMA = "SDMA"
     RDMA = "RDMA"
 
-    def __init__(self, collection_path: str, data: dict, kwargs: dict = None):
+    def __init__(self, collection_path: str, data: dict, **kwargs):
         self.rank_bw_dict = defaultdict(lambda: {
             self.RDMA_TIME_MS: 0,
             self.RDMA_SIZE_MB: 0,
diff --git a/profiler/advisor/dataset/timeline_event_dataset.py b/profiler/advisor/dataset/timeline_event_dataset.py
index ca47227692c3ba2cb1b7956ed119ab62b81ba236..dc4453a7bd75ad29d4ed371c83dd1c50c9bfe874 100644
--- a/profiler/advisor/dataset/timeline_event_dataset.py
+++ b/profiler/advisor/dataset/timeline_event_dataset.py
@@ -78,7 +78,7 @@ class TimelineEventDataset(Dataset):
             return False
 
         if len(self.timeline_data_list) > 1:
-            logger.warning("Please ensure only one trace_view.json in %s, there will analysis first timeline profiling data.", self.timeline_dir)
+            logger.warning("Please ensure only one trace_view.json in %s, there will analyze first timeline profiling data.", self.timeline_dir)
             self.timeline_data_list = [self.timeline_data_list[0]]
 
         result = self.parse_data_with_generator(self._add_event)
diff --git a/profiler/advisor/display/html/render.py b/profiler/advisor/display/html/render.py
index 5b4b17ae4a742cbe087bdba9a98b22077fbf17e0..8ea7c9e0fc22c7da71a673e399fcfc231fbf1453 100644
--- a/profiler/advisor/display/html/render.py
+++ b/profiler/advisor/display/html/render.py
@@ -5,6 +5,7 @@ from typing import List, Dict
 from jinja2 import Environment, FileSystemLoader
 from profiler.advisor.common import constant
 
+from profiler.advisor.config.config import Config
 from profiler.advisor.utils.utils import singleton, safe_write
 
 logger = logging.getLogger()
@@ -41,4 +42,4 @@ class HTMLRender:
             return
 
         safe_write(self.html, save_path)
-        logger.info("Save suggestion to %s.", save_path)
+        logger.info("Save suggestion to %s.", os.path.join(Config().work_path, save_path))
diff --git a/profiler/advisor/interface/interface.py b/profiler/advisor/interface/interface.py
index f9b9fb467ec5f74589c3f381bfd2876587e36b8e..1e0f0e9a563b8512f5db3a7b812a4f05a9c343f0 100644
--- a/profiler/advisor/interface/interface.py
+++ b/profiler/advisor/interface/interface.py
@@ -5,7 +5,7 @@ sys.path.insert(0, os.path.join(os.path.dirname(os.path.dirname(os.path.dirname(
 sys.path.insert(0, os.path.join(os.path.dirname(os.path.dirname(os.path.dirname(os.path.realpath(__file__)))), "compare_tools"))
 
 from profiler.advisor.utils.utils import Timer
-from profiler.advisor.analyzer.computation.profiling_analyzer import ProfilingAnalyzer
+from profiler.advisor.analyzer.computation.profiling_analyzer import AicpuAnalyzer, BlockDimAnalyzer, DynamicShapeAnalyzer, OperatorBoundAnalyzer
 from profiler.advisor.analyzer.schedule.fusion_ops.fusion_ops_analyzer import TimelineFusionOpsAnalyzer
 from profiler.advisor.analyzer.graph_fusion.graph_fusion_analyzer import FusionOPAnalyzer
 from profiler.advisor.common.analyzer_scopes import SupportedScopes
@@ -19,7 +19,10 @@ class Interface:
             SupportedScopes.TIMELINE_FUSION_OPS: TimelineFusionOpsAnalyzer
         }),
         "computation": OrderedDict({
-            SupportedScopes.PROFILING_OPERATOR_ANALYSIS: ProfilingAnalyzer,
+            SupportedScopes.DYNAMIC_SHAPE_ANALYSIS: DynamicShapeAnalyzer,
+            SupportedScopes.AICPU_ANALYSIS: AicpuAnalyzer,
+            SupportedScopes.OPERATOR_NO_BOUND_ANALYSIS: OperatorBoundAnalyzer,
+            SupportedScopes.BLOCK_DIM_ANALYSIS: BlockDimAnalyzer,
             SupportedScopes.GRAPH: FusionOPAnalyzer
         }),
         "communication": OrderedDict(),
diff --git a/profiler/advisor/result/result.py b/profiler/advisor/result/result.py
index 237317051984f0afe13ceb5b6d06bf348934451b..ebee4f2cc3e687d506422ee05d7d4c4614c91b9d 100644
--- a/profiler/advisor/result/result.py
+++ b/profiler/advisor/result/result.py
@@ -139,6 +139,7 @@ class OptimizeResult:
         for sheet_name, sheet_data in self.sheet_recorder.sheet_data.items():
             self.result_writer.add_data(sheet_name, sheet_data.get("headers"), sheet_data.get("data"))
         self.result_writer.save()
+        logger.info("Save problems details file to %s", Config().analysis_result_file)
         self._save_op_file_list()
         TerminalResult().print()
 
diff --git a/profiler/cli/analyze_cli.py b/profiler/cli/analyze_cli.py
index b40da6af3036b19577cdac45ecfd7fe9759b591c..7bd7f1722517edc2e8177d3b88af06a6217cf5f2 100644
--- a/profiler/cli/analyze_cli.py
+++ b/profiler/cli/analyze_cli.py
@@ -42,7 +42,7 @@ def _analyze(dimensions, **kwargs):
 
     for i, (dimension, scope, interface) in enumerate(job_list[::-1]):
         result_list.append(
-            interface.get_result(dimension, scope, render_html=i == len(job_list) - 1, output_dict=False))
+            interface.get_result(dimension, scope, render_html=i == len(job_list) - 1, output_dict=False, **kwargs))
 
     for result in result_list[::-1]:
         if result and hasattr(result, "show"):
diff --git a/profiler/cli/entrance.py b/profiler/cli/entrance.py
index a260553031ecfc904ae8411d944037bdb2f101ab..f668abe651d78bfa3338ac4529189d5354f5c529 100644
--- a/profiler/cli/entrance.py
+++ b/profiler/cli/entrance.py
@@ -65,3 +65,11 @@ msprof_analyze_cli.add_command(analyze_cli, name="advisor")
 msprof_analyze_cli.add_command(compare_cli, name="compare")
 msprof_analyze_cli.add_command(cluster_cli, name="cluster")
 msprof_analyze_cli.add_command(auto_complete_cli, name="auto-completion")
+
+if __name__ == '__main__':
+    msprof_analyze_cli.main(
+        [
+            "advisor", "all", "-d",
+            r"C:\personalC\profiling_data"
+        ]
+    )
\ No newline at end of file