mindstudio-probe 1.1.1__py3-none-any.whl → 1.2.2__py3-none-any.whl

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (226) hide show
  1. {mindstudio_probe-1.1.1.dist-info → mindstudio_probe-1.2.2.dist-info}/METADATA +3 -2
  2. mindstudio_probe-1.2.2.dist-info/RECORD +415 -0
  3. msprobe/CMakeLists.txt +5 -0
  4. msprobe/README.md +16 -21
  5. msprobe/config.json +1 -0
  6. msprobe/core/common/const.py +185 -11
  7. msprobe/core/common/exceptions.py +3 -1
  8. msprobe/core/common/file_utils.py +33 -7
  9. msprobe/core/common/inplace_ops.yaml +4 -0
  10. msprobe/core/common/utils.py +42 -14
  11. msprobe/core/common_config.py +6 -0
  12. msprobe/core/compare/acc_compare.py +139 -128
  13. msprobe/core/compare/check.py +31 -29
  14. msprobe/core/compare/compare_cli.py +17 -16
  15. msprobe/core/compare/highlight.py +186 -99
  16. msprobe/core/compare/layer_mapping/data_scope_parser.py +19 -8
  17. msprobe/core/compare/layer_mapping/layer_mapping.py +21 -14
  18. msprobe/core/compare/layer_mapping/postprocess_pass.py +4 -3
  19. msprobe/core/compare/merge_result/merge_result.py +381 -0
  20. msprobe/core/compare/merge_result/merge_result_cli.py +31 -0
  21. msprobe/core/compare/merge_result/utils.py +81 -0
  22. msprobe/core/compare/multiprocessing_compute.py +2 -2
  23. msprobe/core/compare/npy_compare.py +109 -147
  24. msprobe/core/compare/utils.py +199 -69
  25. msprobe/core/data_dump/data_collector.py +100 -25
  26. msprobe/core/data_dump/data_processor/base.py +130 -28
  27. msprobe/core/data_dump/data_processor/factory.py +8 -3
  28. msprobe/core/data_dump/data_processor/mindspore_processor.py +170 -23
  29. msprobe/core/data_dump/data_processor/pytorch_processor.py +175 -64
  30. msprobe/core/data_dump/json_writer.py +54 -8
  31. msprobe/core/data_dump/scope.py +19 -18
  32. msprobe/core/overflow_check/abnormal_scene.py +9 -5
  33. msprobe/core/overflow_check/checker.py +1 -1
  34. msprobe/core/overflow_check/utils.py +1 -1
  35. msprobe/docs/01.installation.md +121 -17
  36. msprobe/docs/02.config_introduction.md +18 -16
  37. msprobe/docs/03.config_examples.md +24 -0
  38. msprobe/docs/05.data_dump_PyTorch.md +107 -58
  39. msprobe/docs/06.data_dump_MindSpore.md +95 -34
  40. msprobe/docs/07.accuracy_checker_PyTorch.md +18 -18
  41. msprobe/docs/09.accuracy_checker_MindSpore.md +8 -6
  42. msprobe/docs/10.accuracy_compare_PyTorch.md +99 -41
  43. msprobe/docs/11.accuracy_compare_MindSpore.md +249 -48
  44. msprobe/docs/12.overflow_check_PyTorch.md +1 -1
  45. msprobe/docs/19.monitor.md +310 -220
  46. msprobe/docs/21.visualization_PyTorch.md +125 -35
  47. msprobe/docs/22.visualization_MindSpore.md +149 -41
  48. msprobe/docs/23.generate_operator_PyTorch.md +107 -0
  49. msprobe/docs/24.code_mapping_Mindspore.md +28 -0
  50. msprobe/docs/{23.tool_function_introduction.md → 25.tool_function_introduction.md} +1 -0
  51. msprobe/docs/26.data_dump_PyTorch_baseline.md +37 -0
  52. msprobe/docs/27.dump_json_instruction.md +525 -0
  53. msprobe/docs/28.debugger_save_instruction.md +94 -0
  54. msprobe/docs/28.kernel_dump_MindSpore.md +69 -0
  55. msprobe/docs/FAQ.md +26 -2
  56. msprobe/docs/accuracy_checker_MindSpore/accuracy_checker_MindSpore_baseline.md +14 -0
  57. msprobe/docs/data_dump_MindSpore/data_dump_MindSpore_baseline.md +22 -0
  58. msprobe/docs/img/merge_result.png +0 -0
  59. msprobe/docs/img/monitor/step_count_per_record.png +0 -0
  60. msprobe/docs/img/visualization/fuzzy_match_ms.png +0 -0
  61. msprobe/docs/img/visualization/fuzzy_match_pt.png +0 -0
  62. msprobe/docs/img/visualization/tensorboard_1.png +0 -0
  63. msprobe/docs/img/visualization/tensorboard_2.png +0 -0
  64. msprobe/docs/img/visualization/vis_browser_1.png +0 -0
  65. msprobe/docs/img/visualization/vis_browser_2.png +0 -0
  66. msprobe/docs/img/visualization/vis_precision_info.png +0 -0
  67. msprobe/docs/img/visualization/vis_search_info.png +0 -0
  68. msprobe/docs/img/visualization/vis_show_info.png +0 -0
  69. msprobe/docs/img/visualization/vis_showcase.png +0 -0
  70. msprobe/docs/img/visualization/vis_unmatch_info.png +0 -0
  71. msprobe/docs/visualization/GPTModel.png +0 -0
  72. msprobe/docs/visualization/ParallelMLP.png +0 -0
  73. msprobe/docs/visualization/layer_mapping_example.md +132 -0
  74. msprobe/docs/visualization/mapping.png +0 -0
  75. msprobe/docs/visualization/mapping1.png +0 -0
  76. msprobe/docs/visualization/module_name.png +0 -0
  77. msprobe/docs/visualization/module_name1.png +0 -0
  78. msprobe/docs/visualization/no_mapping.png +0 -0
  79. msprobe/docs/visualization/no_mapping1.png +0 -0
  80. msprobe/docs/visualization/no_mapping_analyze.png +0 -0
  81. msprobe/docs/visualization/top_layer.png +0 -0
  82. msprobe/mindspore/__init__.py +11 -0
  83. msprobe/mindspore/api_accuracy_checker/api_accuracy_checker.py +80 -28
  84. msprobe/mindspore/api_accuracy_checker/api_runner.py +54 -16
  85. msprobe/mindspore/api_accuracy_checker/cmd_parser.py +2 -1
  86. msprobe/mindspore/api_accuracy_checker/compute_element.py +52 -8
  87. msprobe/mindspore/api_accuracy_checker/data_manager.py +37 -0
  88. msprobe/mindspore/api_accuracy_checker/main.py +1 -0
  89. msprobe/mindspore/api_accuracy_checker/multi_api_accuracy_checker.py +12 -6
  90. msprobe/mindspore/api_accuracy_checker/multi_data_manager.py +3 -1
  91. msprobe/mindspore/api_accuracy_checker/torch_mindtorch_importer.py +129 -0
  92. msprobe/mindspore/api_accuracy_checker/type_mapping.py +24 -1
  93. msprobe/mindspore/api_accuracy_checker/utils.py +6 -1
  94. msprobe/mindspore/code_mapping/bind.py +264 -0
  95. msprobe/mindspore/code_mapping/cmd_parser.py +40 -0
  96. msprobe/mindspore/code_mapping/graph.py +49 -0
  97. msprobe/mindspore/code_mapping/graph_parser.py +226 -0
  98. msprobe/mindspore/code_mapping/main.py +24 -0
  99. msprobe/mindspore/code_mapping/processor.py +34 -0
  100. msprobe/mindspore/common/const.py +3 -1
  101. msprobe/mindspore/common/utils.py +68 -5
  102. msprobe/mindspore/compare/distributed_compare.py +0 -2
  103. msprobe/mindspore/compare/ms_compare.py +105 -63
  104. msprobe/mindspore/compare/ms_graph_compare.py +14 -5
  105. msprobe/mindspore/debugger/debugger_config.py +28 -2
  106. msprobe/mindspore/debugger/precision_debugger.py +100 -12
  107. msprobe/mindspore/dump/hook_cell/api_registry.py +85 -16
  108. msprobe/mindspore/dump/hook_cell/hook_cell.py +60 -38
  109. msprobe/mindspore/dump/hook_cell/primitive_hooks.py +33 -15
  110. msprobe/mindspore/dump/hook_cell/support_wrap_ops.yaml +11 -1
  111. msprobe/mindspore/dump/hook_cell/wrap_api.py +92 -1
  112. msprobe/mindspore/dump/jit_dump.py +7 -6
  113. msprobe/mindspore/dump/kernel_dump/kernel_config.py +33 -0
  114. msprobe/mindspore/dump/kernel_graph_dump.py +7 -0
  115. msprobe/mindspore/free_benchmark/api_pynative_self_check.py +13 -4
  116. msprobe/mindspore/free_benchmark/perturbation/bit_noise.py +2 -2
  117. msprobe/mindspore/grad_probe/grad_analyzer.py +24 -12
  118. msprobe/mindspore/grad_probe/hook.py +13 -4
  119. msprobe/mindspore/mindtorch/__init__.py +18 -0
  120. msprobe/mindspore/mindtorch/mindtorch_adaptor.py +255 -0
  121. msprobe/mindspore/monitor/anomaly_detect.py +404 -0
  122. msprobe/mindspore/monitor/distributed/__init__.py +0 -0
  123. msprobe/mindspore/monitor/distributed/distributed_ops.yaml +15 -0
  124. msprobe/mindspore/monitor/distributed/stack_blacklist.yaml +5 -0
  125. msprobe/mindspore/monitor/distributed/wrap_distributed.py +300 -0
  126. msprobe/mindspore/monitor/features.py +63 -0
  127. msprobe/mindspore/monitor/module_hook.py +821 -0
  128. msprobe/mindspore/monitor/module_spec_verifier.py +94 -0
  129. msprobe/mindspore/monitor/utils.py +267 -0
  130. msprobe/mindspore/ms_config.py +13 -3
  131. msprobe/mindspore/overflow_check/kernel_graph_overflow_check.py +7 -0
  132. msprobe/mindspore/service.py +347 -107
  133. msprobe/msprobe.py +24 -3
  134. msprobe/pytorch/__init__.py +7 -7
  135. msprobe/pytorch/api_accuracy_checker/common/utils.py +31 -16
  136. msprobe/pytorch/api_accuracy_checker/compare/algorithm.py +41 -8
  137. msprobe/pytorch/api_accuracy_checker/compare/api_precision_compare.py +100 -267
  138. msprobe/pytorch/api_accuracy_checker/compare/api_precision_standard.yaml +4 -1
  139. msprobe/pytorch/api_accuracy_checker/compare/compare.py +69 -68
  140. msprobe/pytorch/api_accuracy_checker/compare/compare_column.py +54 -0
  141. msprobe/pytorch/api_accuracy_checker/compare/compare_input.py +51 -0
  142. msprobe/pytorch/api_accuracy_checker/compare/compare_utils.py +2 -4
  143. msprobe/pytorch/api_accuracy_checker/generate_op_script/op_generator.py +55 -31
  144. msprobe/pytorch/api_accuracy_checker/precision_standard/absolute_threshold.py +106 -0
  145. msprobe/pytorch/api_accuracy_checker/precision_standard/accumulative_error_compare.py +107 -0
  146. msprobe/pytorch/api_accuracy_checker/precision_standard/base_standard.py +151 -0
  147. msprobe/pytorch/api_accuracy_checker/precision_standard/benchmark_compare.py +226 -0
  148. msprobe/pytorch/api_accuracy_checker/precision_standard/binary_consistency.py +68 -0
  149. msprobe/pytorch/api_accuracy_checker/precision_standard/standard_config.py +218 -0
  150. msprobe/pytorch/api_accuracy_checker/precision_standard/standard_register.py +104 -0
  151. msprobe/pytorch/api_accuracy_checker/precision_standard/thousandth_standard.py +63 -0
  152. msprobe/pytorch/api_accuracy_checker/precision_standard/ulp_compare.py +200 -0
  153. msprobe/pytorch/api_accuracy_checker/run_ut/data_generate.py +57 -1
  154. msprobe/pytorch/api_accuracy_checker/run_ut/multi_run_ut.py +2 -1
  155. msprobe/pytorch/api_accuracy_checker/run_ut/run_overflow_check.py +42 -14
  156. msprobe/pytorch/api_accuracy_checker/run_ut/run_ut.py +64 -19
  157. msprobe/pytorch/api_accuracy_checker/run_ut/run_ut_utils.py +34 -4
  158. msprobe/pytorch/api_accuracy_checker/tensor_transport_layer/device_dispatch.py +5 -3
  159. msprobe/pytorch/bench_functions/apply_adam.py +215 -0
  160. msprobe/pytorch/bench_functions/group_norm_silu.py +27 -0
  161. msprobe/pytorch/bench_functions/mish.py +21 -0
  162. msprobe/pytorch/bench_functions/moe_gating_top_k_softmax.py +44 -0
  163. msprobe/pytorch/bench_functions/npu_fusion_attention.py +42 -10
  164. msprobe/pytorch/bench_functions/sort_v2.py +21 -0
  165. msprobe/pytorch/common/parse_json.py +2 -1
  166. msprobe/pytorch/common/utils.py +116 -2
  167. msprobe/pytorch/compare/distributed_compare.py +17 -29
  168. msprobe/pytorch/compare/pt_compare.py +40 -20
  169. msprobe/pytorch/debugger/debugger_config.py +42 -17
  170. msprobe/pytorch/debugger/precision_debugger.py +56 -12
  171. msprobe/pytorch/dump/module_dump/__init__.py +0 -0
  172. msprobe/pytorch/dump/module_dump/module_dump.py +86 -0
  173. msprobe/pytorch/dump/module_dump/module_processer.py +204 -0
  174. msprobe/pytorch/free_benchmark/common/params.py +2 -1
  175. msprobe/pytorch/free_benchmark/common/utils.py +3 -0
  176. msprobe/pytorch/free_benchmark/compare/grad_saver.py +0 -2
  177. msprobe/pytorch/free_benchmark/result_handlers/base_handler.py +31 -47
  178. msprobe/pytorch/free_benchmark/result_handlers/preheat_handler.py +0 -4
  179. msprobe/pytorch/function_factory.py +7 -1
  180. msprobe/pytorch/hook_module/__init__.py +1 -1
  181. msprobe/pytorch/hook_module/hook_module.py +14 -11
  182. msprobe/pytorch/hook_module/register_optimizer_hook.py +59 -0
  183. msprobe/pytorch/hook_module/support_wrap_ops.yaml +36 -1
  184. msprobe/pytorch/hook_module/wrap_distributed.py +10 -8
  185. msprobe/pytorch/hook_module/wrap_functional.py +0 -40
  186. msprobe/pytorch/monitor/anomaly_analyse.py +1 -1
  187. msprobe/pytorch/monitor/anomaly_detect.py +98 -28
  188. msprobe/pytorch/monitor/csv2tb.py +164 -0
  189. msprobe/pytorch/monitor/distributed/wrap_distributed.py +25 -14
  190. msprobe/pytorch/monitor/features.py +3 -3
  191. msprobe/pytorch/monitor/module_hook.py +543 -318
  192. msprobe/pytorch/monitor/module_metric.py +27 -48
  193. msprobe/pytorch/monitor/module_spec_verifier.py +3 -1
  194. msprobe/pytorch/monitor/optimizer_collect.py +76 -56
  195. msprobe/pytorch/monitor/unittest/test_monitor.py +24 -9
  196. msprobe/pytorch/monitor/utils.py +84 -48
  197. msprobe/pytorch/online_dispatch/dispatch.py +8 -2
  198. msprobe/pytorch/parse_tool/lib/compare.py +10 -10
  199. msprobe/pytorch/parse_tool/lib/config.py +5 -7
  200. msprobe/pytorch/parse_tool/lib/file_desc.py +15 -1
  201. msprobe/pytorch/parse_tool/lib/interactive_cli.py +10 -10
  202. msprobe/pytorch/parse_tool/lib/parse_exception.py +7 -7
  203. msprobe/pytorch/parse_tool/lib/parse_tool.py +11 -10
  204. msprobe/pytorch/parse_tool/lib/utils.py +18 -19
  205. msprobe/pytorch/parse_tool/lib/visualization.py +9 -10
  206. msprobe/pytorch/pt_config.py +19 -22
  207. msprobe/pytorch/service.py +264 -115
  208. msprobe/visualization/builder/graph_builder.py +93 -10
  209. msprobe/visualization/builder/msprobe_adapter.py +30 -6
  210. msprobe/visualization/compare/graph_comparator.py +64 -14
  211. msprobe/visualization/compare/mode_adapter.py +1 -15
  212. msprobe/visualization/graph/base_node.py +15 -19
  213. msprobe/visualization/graph/distributed_analyzer.py +395 -0
  214. msprobe/visualization/graph/graph.py +9 -0
  215. msprobe/visualization/graph/node_op.py +4 -2
  216. msprobe/visualization/graph_service.py +100 -27
  217. msprobe/visualization/utils.py +24 -31
  218. mindstudio_probe-1.1.1.dist-info/RECORD +0 -341
  219. msprobe/pytorch/functional/module_dump.py +0 -84
  220. msprobe/pytorch/module_processer.py +0 -150
  221. {mindstudio_probe-1.1.1.dist-info → mindstudio_probe-1.2.2.dist-info}/LICENSE +0 -0
  222. {mindstudio_probe-1.1.1.dist-info → mindstudio_probe-1.2.2.dist-info}/WHEEL +0 -0
  223. {mindstudio_probe-1.1.1.dist-info → mindstudio_probe-1.2.2.dist-info}/entry_points.txt +0 -0
  224. {mindstudio_probe-1.1.1.dist-info → mindstudio_probe-1.2.2.dist-info}/top_level.txt +0 -0
  225. /msprobe/docs/{data_dump_Mindspore → data_dump_MindSpore}/dynamic_graph_quick_start_example.md +0 -0
  226. /msprobe/{pytorch/functional → mindspore/code_mapping}/__init__.py +0 -0
@@ -1,150 +0,0 @@
1
- # Copyright (c) 2024-2024, Huawei Technologies Co., Ltd.
2
- # All rights reserved.
3
- #
4
- # Licensed under the Apache License, Version 2.0 (the "License");
5
- # you may not use this file except in compliance with the License.
6
- # You may obtain a copy of the License at
7
- #
8
- # http://www.apache.org/licenses/LICENSE-2.0
9
- #
10
- # Unless required by applicable law or agreed to in writing, software
11
- # distributed under the License is distributed on an "AS IS" BASIS,
12
- # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13
- # See the License for the specific language governing permissions and
14
- # limitations under the License.
15
-
16
- from functools import wraps
17
-
18
- import torch
19
- from msprobe.core.common.const import Const
20
- from msprobe.core.data_dump.scope import ModuleRangeScope, MixRangeScope
21
- from torch.utils.hooks import BackwardHook
22
-
23
- torch_version_above_or_equal_2 = torch.__version__.split('+')[0] >= '2.0'
24
-
25
-
26
- class ModuleProcesser:
27
- module_count = {}
28
- module_stack = []
29
- api_parent_node = ""
30
- module_node = {}
31
-
32
- def __init__(self, scope):
33
- self.scope = scope if isinstance(scope, (ModuleRangeScope, MixRangeScope)) else None
34
- BackwardHook.setup_input_hook = ModuleProcesser.clone_return_value(BackwardHook.setup_input_hook)
35
- BackwardHook.setup_output_hook = ModuleProcesser.clone_return_value(BackwardHook.setup_output_hook)
36
- BackwardHook.setup_output_hook = ModuleProcesser.filter_tensor_and_tuple(BackwardHook.setup_output_hook)
37
-
38
- @staticmethod
39
- def filter_tensor_and_tuple(func):
40
- @wraps(func)
41
- def wrap_by_filter_tensor_and_tuple(*args, **kwargs):
42
- # setup_output_hook传入非tensor数据,工具后续dump会报错,处理方式是解析非tensor数据的属性,对tensor属性挂hook
43
- # setup_output_hook定义为setup_output_hook(self, args),因此处理第二个位置参数,即*args[1]
44
- if not isinstance(args[1], (torch.Tensor, tuple)):
45
- for item_str in dir(args[1]):
46
- item = getattr(args[1], item_str)
47
- # 处理tensor或者只包含tensor的元组
48
- if isinstance(item, torch.Tensor) or \
49
- (isinstance(item, tuple) and all(isinstance(x, torch.Tensor) for x in item)):
50
- args_new = (args[0], item)
51
- result = func(*args_new, **kwargs)
52
- setattr(args[1], item_str, result)
53
- return args[1]
54
- return func(*args, **kwargs)
55
-
56
- return wrap_by_filter_tensor_and_tuple
57
-
58
- @staticmethod
59
- def clone_return_value(func):
60
- @wraps(func)
61
- def clone_return_value_func(*args, **kwargs):
62
- result = func(*args, **kwargs)
63
- return ModuleProcesser.clone_if_tensor(result)
64
-
65
- return clone_return_value_func
66
-
67
- @staticmethod
68
- def clone_if_tensor(result):
69
- if isinstance(result, torch.Tensor):
70
- return result.clone()
71
- elif isinstance(result, tuple):
72
- return tuple(ModuleProcesser.clone_if_tensor(x) for x in result)
73
- elif isinstance(result, list):
74
- return list(ModuleProcesser.clone_if_tensor(x) for x in result)
75
- elif isinstance(result, dict):
76
- return {k: ModuleProcesser.clone_if_tensor(v) for k, v in result.items()}
77
- else:
78
- return result
79
-
80
- @staticmethod
81
- def module_count_func(module_name):
82
- if module_name not in ModuleProcesser.module_count:
83
- ModuleProcesser.module_count[module_name] = 0
84
- else:
85
- ModuleProcesser.module_count[module_name] += 1
86
- return ModuleProcesser.module_count[module_name]
87
-
88
- @classmethod
89
- def reset_module_stats(cls):
90
- cls.module_count = {}
91
- cls.module_stack = []
92
- cls.api_parent_node = ""
93
- cls.module_node = {}
94
-
95
- def node_hook(self, name_prefix, start_or_stop, **kwargs):
96
-
97
- def pre_hook(module, input, output=None):
98
- try:
99
- index = ModuleProcesser.module_count_func(name_prefix)
100
- except IndexError as e:
101
- index = None
102
- pass
103
- module.mindstudio_reserved_name = full_name = name_prefix + Const.SEP + str(index)
104
- if self.module_stack:
105
- ModuleProcesser.module_node[full_name] = self.module_stack[-1]
106
- else:
107
- ModuleProcesser.module_node[full_name] = None
108
-
109
- ModuleProcesser.module_stack.append(full_name)
110
- if self.module_stack:
111
- ModuleProcesser.api_parent_node = self.module_stack[-1]
112
- if self.scope:
113
- self.scope.begin_module(full_name)
114
-
115
- def end_hook(module, input, output=None):
116
- if self.module_stack:
117
- ModuleProcesser.module_stack.pop()
118
- if self.module_stack:
119
- ModuleProcesser.api_parent_node = self.module_stack[-1]
120
- else:
121
- ModuleProcesser.api_parent_node = None
122
- if self.scope:
123
- self.scope.end_module(module.mindstudio_reserved_name)
124
-
125
- def backward_hook(module, input, output=None):
126
- try:
127
- index = ModuleProcesser.module_count_func(name_prefix)
128
- except IndexError as e:
129
- index = None
130
- pass
131
- module.mindstudio_reserved_name = full_name = name_prefix + Const.SEP + str(index)
132
- forward_full_name = full_name.replace(Const.BACKWARD, Const.FORWARD)
133
- ModuleProcesser.module_node[full_name] = ModuleProcesser.module_node[forward_full_name].replace(
134
- Const.FORWARD, Const.BACKWARD) if ModuleProcesser.module_node[forward_full_name] else None
135
- ModuleProcesser.api_parent_node = None
136
- if self.scope:
137
- self.scope.begin_module(full_name)
138
-
139
- if torch_version_above_or_equal_2:
140
- if Const.START in start_or_stop:
141
- return pre_hook
142
- else:
143
- return end_hook
144
- else:
145
- if Const.FORWARD in name_prefix and Const.START in start_or_stop:
146
- return pre_hook
147
- elif Const.BACKWARD in name_prefix:
148
- return backward_hook
149
- else:
150
- return end_hook