mindstudio-probe 8.3.3__py3-none-any.whl → 26.0.0a1__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.
- {mindstudio_probe-8.3.3.dist-info → mindstudio_probe-26.0.0a1.dist-info}/METADATA +26 -14
- mindstudio_probe-26.0.0a1.dist-info/RECORD +498 -0
- {mindstudio_probe-8.3.3.dist-info → mindstudio_probe-26.0.0a1.dist-info}/WHEEL +1 -1
- mindstudio_probe-26.0.0a1.dist-info/entry_points.txt +5 -0
- mindstudio_probe-26.0.0a1.dist-info/licenses/LICENSE +124 -0
- mindstudio_probe-26.0.0a1.dist-info/top_level.txt +2 -0
- msprobe/__init__.py +12 -13
- msprobe/config.json +9 -31
- msprobe/core/__init__.py +12 -11
- msprobe/core/acc_check/acc_check_cli.py +145 -0
- msprobe/core/common/const.py +97 -38
- msprobe/core/common/db_manager.py +133 -12
- msprobe/core/common/decorator.py +12 -11
- msprobe/core/common/exceptions.py +12 -11
- msprobe/core/common/file_utils.py +101 -25
- msprobe/core/common/framework_adapter.py +36 -25
- msprobe/core/common/global_lock.py +12 -11
- msprobe/core/common/inplace_op_checker.py +12 -11
- msprobe/core/common/log.py +22 -11
- msprobe/core/common/megatron_utils.py +566 -11
- msprobe/core/common/parallel_state.py +12 -11
- msprobe/core/common/runtime.py +12 -11
- msprobe/core/common/utils.py +41 -41
- msprobe/core/compare/acc_compare.py +361 -104
- msprobe/core/compare/atb_data_compare.py +422 -0
- msprobe/core/compare/auto_compare.py +134 -0
- msprobe/core/compare/check.py +14 -17
- msprobe/core/compare/compare_cli.py +72 -149
- msprobe/core/compare/config.py +12 -13
- msprobe/core/compare/diff_analyze/first_diff_analyze.py +28 -15
- msprobe/core/compare/diff_analyze/ignore_op_list.yaml +3 -0
- msprobe/core/compare/find_first/analyzer.py +18 -18
- msprobe/core/compare/find_first/graph.py +12 -11
- msprobe/core/compare/find_first/utils.py +13 -12
- msprobe/core/compare/indicator_analysis/__init__.py +15 -0
- msprobe/core/compare/indicator_analysis/algorithm.py +363 -0
- msprobe/core/compare/indicator_analysis/api_data.py +141 -0
- msprobe/core/compare/indicator_analysis/calculator.py +181 -0
- msprobe/core/compare/indicator_analysis/utils.py +116 -0
- msprobe/core/compare/layer_mapping/__init__.py +12 -11
- msprobe/core/compare/layer_mapping/data_scope_parser.py +20 -11
- msprobe/core/compare/layer_mapping/layer_mapping.py +14 -13
- msprobe/core/compare/layer_mapping/postprocess_pass.py +13 -11
- msprobe/core/compare/merge_result/merge_result.py +12 -11
- msprobe/core/compare/merge_result/merge_result_cli.py +12 -11
- msprobe/core/compare/merge_result/utils.py +12 -11
- msprobe/core/compare/multiprocessing_compute.py +13 -14
- msprobe/core/compare/npy_compare.py +13 -11
- msprobe/core/compare/offline_data_compare.py +160 -0
- msprobe/core/compare/stats_diff_calc.py +39 -0
- msprobe/core/compare/torchair_acc_cmp.py +764 -0
- msprobe/core/compare/torchair_cmp_utils.py +338 -0
- msprobe/core/compare/utils.py +140 -49
- msprobe/core/config_check/__init__.py +12 -11
- msprobe/core/config_check/checkers/__init__.py +12 -11
- msprobe/core/config_check/checkers/base_checker.py +15 -14
- msprobe/core/config_check/checkers/dataset_checker.py +13 -12
- msprobe/core/config_check/checkers/env_args_checker.py +13 -12
- msprobe/core/config_check/checkers/hyperparameter_checker.py +16 -15
- msprobe/core/config_check/checkers/pip_checker.py +15 -15
- msprobe/core/config_check/checkers/random_checker.py +13 -12
- msprobe/core/config_check/checkers/weights_checker.py +14 -12
- msprobe/core/config_check/ckpt_compare/ckpt_comparator.py +13 -17
- msprobe/core/config_check/ckpt_compare/megatron_loader.py +13 -12
- msprobe/core/config_check/ckpt_compare/metrics.py +12 -11
- msprobe/core/config_check/config_check_cli.py +18 -17
- msprobe/core/config_check/config_checker.py +16 -14
- msprobe/core/config_check/resource/dependency.yaml +15 -12
- msprobe/core/config_check/resource/env.yaml +12 -11
- msprobe/core/config_check/utils/hyperparameter_parser.py +12 -11
- msprobe/core/config_check/utils/utils.py +12 -11
- msprobe/core/{data_dump → dump/api_dump}/api_registry.py +12 -11
- msprobe/core/{common_config.py → dump/common_config.py} +13 -24
- msprobe/core/dump/data_dump/data_collector.py +257 -0
- msprobe/core/{data_dump → dump/data_dump}/data_processor/base.py +45 -36
- msprobe/core/{data_dump → dump/data_dump}/data_processor/factory.py +33 -25
- msprobe/core/{data_dump → dump/data_dump}/data_processor/mindspore_processor.py +37 -113
- msprobe/core/{data_dump → dump/data_dump}/data_processor/pytorch_processor.py +364 -131
- msprobe/core/{data_dump → dump/data_dump}/json_writer.py +24 -31
- msprobe/core/{data_dump → dump/data_dump}/scope.py +12 -13
- msprobe/core/{debugger → dump/debugger}/precision_debugger.py +15 -23
- msprobe/core/dump/dump2db/db_utils.py +215 -0
- msprobe/core/dump/dump2db/dump2db.py +409 -0
- msprobe/core/{hook_manager.py → dump/hook_manager.py} +38 -87
- msprobe/core/dump/kernel_dump/kernel_config.py +34 -0
- msprobe/core/{service.py → dump/service.py} +43 -27
- msprobe/core/install_deps/install_deps.py +51 -0
- msprobe/core/monitor/anomaly_processor.py +13 -11
- msprobe/core/monitor/csv2db.py +73 -93
- msprobe/core/monitor/db_utils.py +140 -205
- msprobe/core/monitor/utils.py +18 -17
- msprobe/core/monitor_v2/__init__.py +20 -0
- msprobe/core/monitor_v2/base.py +83 -0
- msprobe/core/monitor_v2/cc.py +287 -0
- msprobe/core/monitor_v2/factory.py +81 -0
- msprobe/core/monitor_v2/module.py +201 -0
- msprobe/core/monitor_v2/optimizer.py +245 -0
- msprobe/core/monitor_v2/param.py +154 -0
- msprobe/core/monitor_v2/trainer.py +326 -0
- msprobe/core/monitor_v2/utils.py +122 -0
- msprobe/core/monitor_v2/weight_grad.py +419 -0
- msprobe/core/monitor_v2/writer.py +162 -0
- msprobe/core/overflow_check/abnormal_scene.py +12 -11
- msprobe/core/overflow_check/api_info.py +12 -11
- msprobe/core/overflow_check/checker.py +12 -11
- msprobe/core/overflow_check/filter.py +13 -11
- msprobe/core/overflow_check/level.py +12 -11
- msprobe/core/overflow_check/utils.py +12 -11
- msprobe/core/single_save/single_comparator.py +12 -11
- msprobe/core/single_save/single_saver.py +12 -11
- msprobe/infer/__init__.py +16 -0
- msprobe/infer/offline/__init__.py +16 -0
- msprobe/infer/offline/compare/__init__.py +16 -0
- msprobe/infer/offline/compare/msquickcmp/__init__.py +16 -0
- msprobe/infer/offline/compare/msquickcmp/adapter_cli/__init__.py +16 -0
- msprobe/infer/offline/compare/msquickcmp/adapter_cli/args_adapter.py +46 -0
- msprobe/infer/offline/compare/msquickcmp/atc/__init__.py +16 -0
- msprobe/infer/offline/compare/msquickcmp/atc/atc_utils.py +98 -0
- msprobe/infer/offline/compare/msquickcmp/cmp_process.py +328 -0
- msprobe/infer/offline/compare/msquickcmp/common/__init__.py +16 -0
- msprobe/infer/offline/compare/msquickcmp/common/args_check.py +112 -0
- msprobe/infer/offline/compare/msquickcmp/common/convert.py +74 -0
- msprobe/infer/offline/compare/msquickcmp/common/dump_data.py +121 -0
- msprobe/infer/offline/compare/msquickcmp/common/dynamic_argument_bean.py +39 -0
- msprobe/infer/offline/compare/msquickcmp/common/utils.py +669 -0
- msprobe/infer/offline/compare/msquickcmp/config.ini +6 -0
- msprobe/infer/offline/compare/msquickcmp/dump/__init__.py +16 -0
- msprobe/infer/offline/compare/msquickcmp/dump/args_adapter.py +50 -0
- msprobe/infer/offline/compare/msquickcmp/dump/dump_process.py +91 -0
- msprobe/infer/offline/compare/msquickcmp/install_aclruntime_aisbench.sh +180 -0
- msprobe/infer/offline/compare/msquickcmp/main.py +199 -0
- msprobe/infer/offline/compare/msquickcmp/net_compare/__init__.py +16 -0
- msprobe/infer/offline/compare/msquickcmp/net_compare/net_compare.py +277 -0
- msprobe/infer/offline/compare/msquickcmp/npu/__init__.py +16 -0
- msprobe/infer/offline/compare/msquickcmp/npu/npu_dump_data.py +558 -0
- msprobe/infer/offline/compare/msquickcmp/npu/om_parser.py +416 -0
- msprobe/infer/offline/compare/msquickcmp/onnx_model/__init__.py +16 -0
- msprobe/infer/offline/compare/msquickcmp/onnx_model/onnx_dump_data.py +374 -0
- msprobe/infer/utils/__init__.py +15 -0
- msprobe/infer/utils/acc_cmp.py +94 -0
- msprobe/infer/utils/check/__init__.py +37 -0
- msprobe/infer/utils/check/args_checker.py +35 -0
- msprobe/infer/utils/check/checker.py +227 -0
- msprobe/infer/utils/check/dict_checker.py +78 -0
- msprobe/infer/utils/check/func_wrapper.py +96 -0
- msprobe/infer/utils/check/list_checker.py +56 -0
- msprobe/infer/utils/check/number_checker.py +64 -0
- msprobe/infer/utils/check/obj_checker.py +41 -0
- msprobe/infer/utils/check/path_checker.py +249 -0
- msprobe/infer/utils/check/rule.py +126 -0
- msprobe/infer/utils/check/string_checker.py +66 -0
- msprobe/infer/utils/cmp_algorithm.py +261 -0
- msprobe/infer/utils/constants.py +112 -0
- msprobe/infer/utils/file_open_check.py +337 -0
- msprobe/infer/utils/util.py +177 -0
- msprobe/mindspore/__init__.py +14 -13
- msprobe/mindspore/api_accuracy_checker/api_accuracy_checker.py +14 -13
- msprobe/mindspore/api_accuracy_checker/api_info.py +12 -11
- msprobe/mindspore/api_accuracy_checker/api_runner.py +12 -11
- msprobe/mindspore/api_accuracy_checker/base_compare_algorithm.py +12 -11
- msprobe/mindspore/api_accuracy_checker/bench_functions/flash_attention_score.py +12 -11
- msprobe/mindspore/api_accuracy_checker/bench_functions/fusion_operator.py +12 -11
- msprobe/mindspore/api_accuracy_checker/checker_support_api.yaml +12 -11
- msprobe/mindspore/api_accuracy_checker/cmd_parser.py +15 -14
- msprobe/mindspore/api_accuracy_checker/compute_element.py +12 -11
- msprobe/mindspore/api_accuracy_checker/data_manager.py +13 -11
- msprobe/mindspore/api_accuracy_checker/main.py +12 -11
- msprobe/mindspore/api_accuracy_checker/multi_api_accuracy_checker.py +14 -12
- msprobe/mindspore/api_accuracy_checker/multi_data_manager.py +13 -11
- msprobe/mindspore/api_accuracy_checker/torch_mindtorch_importer.py +12 -11
- msprobe/mindspore/api_accuracy_checker/type_mapping.py +12 -11
- msprobe/mindspore/api_accuracy_checker/utils.py +12 -11
- msprobe/mindspore/common/const.py +15 -74
- msprobe/mindspore/common/log.py +12 -11
- msprobe/mindspore/common/utils.py +30 -15
- msprobe/mindspore/compare/common_dir_compare.py +21 -23
- msprobe/mindspore/compare/distributed_compare.py +18 -16
- msprobe/mindspore/compare/ms_compare.py +14 -14
- msprobe/mindspore/compare/ms_graph_compare.py +26 -20
- msprobe/mindspore/compare/utils.py +14 -12
- msprobe/mindspore/{cell_processor.py → dump/cell_processor.py} +15 -14
- msprobe/mindspore/{debugger → dump/debugger}/debugger_config.py +12 -30
- msprobe/mindspore/{debugger → dump/debugger}/precision_debugger.py +43 -45
- msprobe/mindspore/dump/{cell_dump_process.py → dump_processor/cell_dump_process.py} +31 -17
- msprobe/mindspore/dump/{cell_dump_with_insert_gradient.py → dump_processor/cell_dump_with_insert_gradient.py} +18 -14
- msprobe/mindspore/dump/{dump_tool_factory.py → dump_processor/dump_tool_factory.py} +16 -15
- msprobe/mindspore/dump/{graph_mode_cell_dump.py → dump_processor/graph_mode_cell_dump.py} +16 -15
- msprobe/mindspore/dump/{graph_tensor_dump.py → dump_processor/graph_tensor_dump.py} +134 -133
- msprobe/mindspore/dump/{hook_cell → dump_processor/hook_cell}/api_register.py +15 -14
- msprobe/mindspore/dump/{hook_cell → dump_processor/hook_cell}/hook_cell.py +12 -11
- msprobe/mindspore/dump/{hook_cell → dump_processor/hook_cell}/ms_hook_manager.py +47 -20
- msprobe/mindspore/dump/{hook_cell → dump_processor/hook_cell}/primitive_hooks.py +14 -13
- msprobe/mindspore/dump/{hook_cell → dump_processor/hook_cell}/support_wrap_ops.yaml +13 -11
- msprobe/mindspore/dump/{jit_dump.py → dump_processor/jit_dump.py} +14 -13
- msprobe/mindspore/dump/{kernel_graph_dump.py → dump_processor/kernel_graph_dump.py} +13 -12
- msprobe/mindspore/dump/{kernel_kbyk_dump.py → dump_processor/kernel_kbyk_dump.py} +13 -12
- msprobe/mindspore/{exception_dump → dump/exception_dump}/exception_dump_tool_factory.py +14 -13
- msprobe/mindspore/{exception_dump → dump/exception_dump}/kernel_graph_exception_dump.py +13 -12
- msprobe/mindspore/{mindspore_service.py → dump/mindspore_service.py} +18 -17
- msprobe/mindspore/dump/mindtorch/__init__.py +19 -0
- msprobe/mindspore/dump/ms_config.py +105 -0
- msprobe/mindspore/{overflow_check → dump/overflow_check}/kernel_graph_overflow_check.py +13 -12
- msprobe/mindspore/{overflow_check → dump/overflow_check}/overflow_check_tool_factory.py +14 -13
- msprobe/mindspore/dump/task_handler_factory.py +43 -0
- msprobe/mindspore/monitor/common_func.py +12 -11
- msprobe/mindspore/monitor/data_writers.py +12 -11
- msprobe/mindspore/monitor/distributed/wrap_distributed.py +93 -39
- msprobe/mindspore/monitor/features.py +12 -11
- msprobe/mindspore/monitor/module_hook.py +19 -22
- msprobe/mindspore/monitor/optimizer_collect.py +29 -25
- msprobe/mindspore/monitor/utils.py +13 -11
- msprobe/msaccucmp/advisor/__init__.py +16 -0
- msprobe/msaccucmp/advisor/advisor_const.py +65 -0
- msprobe/msaccucmp/advisor/advisor_result.py +73 -0
- msprobe/msaccucmp/advisor/compare_advisor.py +99 -0
- msprobe/msaccucmp/advisor/input_advisor.py +66 -0
- msprobe/msaccucmp/advisor/node_advisor.py +68 -0
- msprobe/msaccucmp/advisor/overflow_advisor.py +58 -0
- msprobe/msaccucmp/algorithm_manager/__init__.py +16 -0
- msprobe/msaccucmp/algorithm_manager/algorithm_manager.py +464 -0
- msprobe/msaccucmp/algorithm_manager/algorithm_parameter.py +42 -0
- msprobe/msaccucmp/algorithm_manager/builtin_algorithm/alg_AccumulatedRelativeError.py +46 -0
- msprobe/msaccucmp/algorithm_manager/builtin_algorithm/alg_CosineSimilarity.py +58 -0
- msprobe/msaccucmp/algorithm_manager/builtin_algorithm/alg_KullbackLeiblerDivergence.py +84 -0
- msprobe/msaccucmp/algorithm_manager/builtin_algorithm/alg_MaxAbsoluteError.py +41 -0
- msprobe/msaccucmp/algorithm_manager/builtin_algorithm/alg_MaxRelativeError.py +46 -0
- msprobe/msaccucmp/algorithm_manager/builtin_algorithm/alg_MeanAbsoluteError.py +41 -0
- msprobe/msaccucmp/algorithm_manager/builtin_algorithm/alg_MeanRelativeError.py +46 -0
- msprobe/msaccucmp/algorithm_manager/builtin_algorithm/alg_RelativeEuclideanDistance.py +46 -0
- msprobe/msaccucmp/algorithm_manager/builtin_algorithm/alg_RootMeanSquareError.py +40 -0
- msprobe/msaccucmp/algorithm_manager/builtin_algorithm/alg_StandardDeviation.py +47 -0
- msprobe/msaccucmp/cmp_utils/__init__.py +16 -0
- msprobe/msaccucmp/cmp_utils/common.py +113 -0
- msprobe/msaccucmp/cmp_utils/constant/__init__.py +16 -0
- msprobe/msaccucmp/cmp_utils/constant/compare_error.py +81 -0
- msprobe/msaccucmp/cmp_utils/constant/const_manager.py +530 -0
- msprobe/msaccucmp/cmp_utils/file_utils.py +497 -0
- msprobe/msaccucmp/cmp_utils/log.py +257 -0
- msprobe/msaccucmp/cmp_utils/multi_process/__init__.py +16 -0
- msprobe/msaccucmp/cmp_utils/multi_process/multi_convert_process.py +140 -0
- msprobe/msaccucmp/cmp_utils/multi_process/progress.py +78 -0
- msprobe/msaccucmp/cmp_utils/path_check.py +274 -0
- msprobe/msaccucmp/cmp_utils/reg_manager.py +98 -0
- msprobe/msaccucmp/cmp_utils/tlv_parse.py +279 -0
- msprobe/msaccucmp/cmp_utils/utils.py +356 -0
- msprobe/msaccucmp/cmp_utils/utils_type.py +63 -0
- msprobe/msaccucmp/compare_vector.py +48 -0
- msprobe/msaccucmp/conversion/__init__.py +16 -0
- msprobe/msaccucmp/conversion/data_conversion.py +277 -0
- msprobe/msaccucmp/conversion/dtype_conversion.py +99 -0
- msprobe/msaccucmp/conversion/shape_format_conversion.py +477 -0
- msprobe/msaccucmp/conversion/tensor_conversion.py +369 -0
- msprobe/msaccucmp/dump_data_conversion.py +46 -0
- msprobe/msaccucmp/dump_parse/__init__.py +16 -0
- msprobe/msaccucmp/dump_parse/big_dump_data.py +317 -0
- msprobe/msaccucmp/dump_parse/dump.py +423 -0
- msprobe/msaccucmp/dump_parse/dump_data_object.py +322 -0
- msprobe/msaccucmp/dump_parse/dump_data_parser.py +436 -0
- msprobe/msaccucmp/dump_parse/dump_utils.py +246 -0
- msprobe/msaccucmp/dump_parse/ffts_parser.py +137 -0
- msprobe/msaccucmp/dump_parse/mapping.py +62 -0
- msprobe/msaccucmp/dump_parse/nano_dump_data.py +392 -0
- msprobe/msaccucmp/dump_parse/proto_dump_data.py +308 -0
- msprobe/msaccucmp/dump_parser.py +90 -0
- msprobe/msaccucmp/format_manager/__init__.py +16 -0
- msprobe/msaccucmp/format_manager/builtin_format_convert/convert_FRACTAL_NZ_to_NCHW.py +53 -0
- msprobe/msaccucmp/format_manager/builtin_format_convert/convert_FRACTAL_NZ_to_ND.py +52 -0
- msprobe/msaccucmp/format_manager/builtin_format_convert/convert_FRACTAL_NZ_to_NHWC.py +53 -0
- msprobe/msaccucmp/format_manager/builtin_format_convert/convert_FRACTAL_Z_to_HWCN.py +47 -0
- msprobe/msaccucmp/format_manager/builtin_format_convert/convert_FRACTAL_Z_to_NCHW.py +47 -0
- msprobe/msaccucmp/format_manager/builtin_format_convert/convert_HWCN_to_FRACTAL_Z.py +89 -0
- msprobe/msaccucmp/format_manager/builtin_format_convert/convert_HWCN_to_NCHW.py +37 -0
- msprobe/msaccucmp/format_manager/builtin_format_convert/convert_HWCN_to_NHWC.py +37 -0
- msprobe/msaccucmp/format_manager/builtin_format_convert/convert_NC1HWC0_to_HWCN.py +43 -0
- msprobe/msaccucmp/format_manager/builtin_format_convert/convert_NC1HWC0_to_NCHW.py +48 -0
- msprobe/msaccucmp/format_manager/builtin_format_convert/convert_NC1HWC0_to_NHWC.py +43 -0
- msprobe/msaccucmp/format_manager/builtin_format_convert/convert_NCHW_to_FRACTAL_Z.py +87 -0
- msprobe/msaccucmp/format_manager/builtin_format_convert/convert_NCHW_to_NHWC.py +37 -0
- msprobe/msaccucmp/format_manager/builtin_format_convert/convert_NDC1HWC0_to_NCDHW.py +48 -0
- msprobe/msaccucmp/format_manager/builtin_format_convert/convert_NDC1HWC0_to_ND.py +44 -0
- msprobe/msaccucmp/format_manager/builtin_format_convert/convert_NHWC_to_FRACTAL_Z.py +87 -0
- msprobe/msaccucmp/format_manager/builtin_format_convert/convert_NHWC_to_HWCN.py +37 -0
- msprobe/msaccucmp/format_manager/builtin_format_convert/convert_NHWC_to_NCHW.py +37 -0
- msprobe/msaccucmp/format_manager/format_manager.py +307 -0
- msprobe/msaccucmp/inplace_layer_process.py +186 -0
- msprobe/msaccucmp/msaccucmp.py +532 -0
- msprobe/msaccucmp/mscmp_advisor.py +128 -0
- msprobe/msaccucmp/overflow/__init__.py +16 -0
- msprobe/msaccucmp/overflow/overflow_analyse.py +305 -0
- msprobe/msaccucmp/overflow/overflow_detection.py +143 -0
- msprobe/msaccucmp/pytorch_cmp/__init__.py +16 -0
- msprobe/msaccucmp/pytorch_cmp/compare_pytorch.py +389 -0
- msprobe/msaccucmp/pytorch_cmp/hdf5_parser.py +377 -0
- msprobe/msaccucmp/pytorch_cmp/pytorch_dump_data.py +461 -0
- msprobe/msaccucmp/shape_conversion.py +41 -0
- msprobe/msaccucmp/vector_cmp/__init__.py +16 -0
- msprobe/msaccucmp/vector_cmp/batch_compare.py +197 -0
- msprobe/msaccucmp/vector_cmp/compare_detail/__init__.py +16 -0
- msprobe/msaccucmp/vector_cmp/compare_detail/compare_detail.py +245 -0
- msprobe/msaccucmp/vector_cmp/compare_detail/detail.py +182 -0
- msprobe/msaccucmp/vector_cmp/compare_detail/detail_writer.py +580 -0
- msprobe/msaccucmp/vector_cmp/fusion_manager/__init__.py +16 -0
- msprobe/msaccucmp/vector_cmp/fusion_manager/compare_fusion_op.py +588 -0
- msprobe/msaccucmp/vector_cmp/fusion_manager/compare_npu_vs_npu.py +339 -0
- msprobe/msaccucmp/vector_cmp/fusion_manager/compare_result.py +326 -0
- msprobe/msaccucmp/vector_cmp/fusion_manager/compare_rule.py +156 -0
- msprobe/msaccucmp/vector_cmp/fusion_manager/fusion_op.py +204 -0
- msprobe/msaccucmp/vector_cmp/fusion_manager/fusion_rule_parser.py +635 -0
- msprobe/msaccucmp/vector_cmp/fusion_manager/quant_filter.py +187 -0
- msprobe/msaccucmp/vector_cmp/range_manager/__init__.py +16 -0
- msprobe/msaccucmp/vector_cmp/range_manager/range_manager.py +100 -0
- msprobe/msaccucmp/vector_cmp/range_manager/range_mode.py +94 -0
- msprobe/msaccucmp/vector_cmp/range_manager/select_mode.py +86 -0
- msprobe/msaccucmp/vector_cmp/vector_comparison.py +535 -0
- msprobe/msprobe.py +101 -130
- msprobe/overflow_check/__init__.py +15 -0
- msprobe/{nan_analyze → overflow_check}/analyzer.py +38 -27
- msprobe/{nan_analyze → overflow_check}/graph.py +28 -27
- msprobe/{nan_analyze → overflow_check}/utils.py +15 -14
- msprobe/pytorch/__init__.py +20 -14
- msprobe/pytorch/aclgraph_dump/__init__.py +45 -0
- msprobe/pytorch/aclgraph_dump/_meta.py +26 -0
- msprobe/pytorch/api_accuracy_checker/{run_ut/run_ut.py → acc_check/acc_check.py} +50 -45
- msprobe/pytorch/api_accuracy_checker/{run_ut/run_ut_utils.py → acc_check/acc_check_utils.py} +201 -30
- msprobe/pytorch/api_accuracy_checker/{run_ut → acc_check}/data_generate.py +56 -16
- msprobe/pytorch/api_accuracy_checker/{run_ut/multi_run_ut.py → acc_check/multi_acc_check.py} +32 -47
- msprobe/pytorch/api_accuracy_checker/{run_ut → acc_check}/run_overflow_check.py +19 -18
- msprobe/pytorch/api_accuracy_checker/common/config.py +22 -20
- msprobe/pytorch/api_accuracy_checker/common/utils.py +72 -13
- msprobe/pytorch/api_accuracy_checker/compare/algorithm.py +41 -11
- msprobe/pytorch/api_accuracy_checker/compare/api_precision_compare.py +23 -14
- msprobe/pytorch/api_accuracy_checker/compare/compare.py +45 -32
- msprobe/pytorch/api_accuracy_checker/compare/compare_column.py +12 -11
- msprobe/pytorch/api_accuracy_checker/compare/compare_input.py +14 -12
- msprobe/pytorch/api_accuracy_checker/compare/compare_utils.py +14 -12
- msprobe/pytorch/api_accuracy_checker/precision_standard/absolute_threshold.py +12 -11
- msprobe/pytorch/api_accuracy_checker/precision_standard/accumulative_error_compare.py +12 -11
- msprobe/pytorch/api_accuracy_checker/precision_standard/base_standard.py +21 -19
- msprobe/pytorch/api_accuracy_checker/precision_standard/benchmark_compare.py +14 -13
- msprobe/pytorch/api_accuracy_checker/precision_standard/binary_consistency.py +12 -11
- msprobe/pytorch/api_accuracy_checker/precision_standard/standard_config.py +60 -11
- msprobe/pytorch/api_accuracy_checker/precision_standard/standard_register.py +27 -16
- msprobe/pytorch/api_accuracy_checker/precision_standard/thousandth_standard.py +13 -11
- msprobe/pytorch/api_accuracy_checker/precision_standard/ulp_compare.py +39 -18
- msprobe/pytorch/bench_functions/__init__.py +12 -11
- msprobe/pytorch/bench_functions/apply_adam.py +12 -11
- msprobe/pytorch/bench_functions/apply_adam_w.py +12 -11
- msprobe/pytorch/bench_functions/confusion_transpose.py +12 -11
- msprobe/pytorch/bench_functions/fast_gelu.py +12 -11
- msprobe/pytorch/bench_functions/group_norm_silu.py +12 -11
- msprobe/pytorch/bench_functions/layer_norm_eval.py +12 -11
- msprobe/pytorch/bench_functions/linear.py +12 -11
- msprobe/pytorch/bench_functions/matmul_backward.py +12 -11
- msprobe/pytorch/bench_functions/mish.py +12 -11
- msprobe/pytorch/bench_functions/moe_gating_top_k_softmax.py +12 -11
- msprobe/pytorch/bench_functions/npu_fusion_attention.py +12 -11
- msprobe/pytorch/bench_functions/rms_norm.py +12 -11
- msprobe/pytorch/bench_functions/rotary_mul.py +12 -11
- msprobe/pytorch/bench_functions/scaled_mask_softmax.py +12 -11
- msprobe/pytorch/bench_functions/sort_v2.py +12 -11
- msprobe/pytorch/bench_functions/swiglu.py +12 -11
- msprobe/pytorch/common/__init__.py +12 -11
- msprobe/pytorch/common/log.py +12 -11
- msprobe/pytorch/common/parse_json.py +12 -11
- msprobe/pytorch/common/utils.py +52 -19
- msprobe/pytorch/compare/distributed_compare.py +13 -13
- msprobe/pytorch/compare/match.py +12 -11
- msprobe/pytorch/compare/pt_compare.py +14 -20
- msprobe/pytorch/compare/pt_diff_analyze.py +12 -11
- msprobe/pytorch/compare/utils.py +12 -11
- msprobe/pytorch/{hook_module → dump/api_dump}/api_register.py +18 -16
- msprobe/pytorch/{hook_module → dump/api_dump}/hook_module.py +14 -13
- msprobe/pytorch/{hook_module → dump/api_dump}/pt_hook_manager.py +68 -23
- msprobe/pytorch/{hook_module → dump/api_dump}/register_optimizer_hook.py +13 -11
- msprobe/pytorch/{hook_module → dump/api_dump}/script_wrapper.py +17 -14
- msprobe/pytorch/{hook_module → dump/api_dump}/utils.py +12 -11
- msprobe/pytorch/{debugger → dump/debugger}/debugger_config.py +23 -38
- msprobe/pytorch/dump/debugger/precision_debugger.py +130 -0
- msprobe/pytorch/{function_factory.py → dump/function_factory.py} +12 -11
- msprobe/pytorch/dump/module_dump/hook_wrapper.py +17 -13
- msprobe/pytorch/dump/module_dump/module_dump.py +16 -15
- msprobe/pytorch/dump/module_dump/{module_processer.py → module_processor.py} +54 -42
- msprobe/pytorch/dump/pt_config.py +128 -0
- msprobe/pytorch/{pytorch_service.py → dump/pytorch_service.py} +22 -21
- msprobe/pytorch/monitor/csv2tb.py +13 -11
- msprobe/pytorch/monitor/data_writers.py +13 -11
- msprobe/pytorch/monitor/distributed/wrap_distributed.py +13 -11
- msprobe/pytorch/monitor/features.py +12 -11
- msprobe/pytorch/monitor/module_hook.py +67 -59
- msprobe/pytorch/monitor/module_metric.py +13 -11
- msprobe/pytorch/monitor/optimizer_collect.py +37 -35
- msprobe/pytorch/monitor/utils.py +13 -11
- msprobe/pytorch/monitor/visualizer.py +12 -11
- msprobe/pytorch/torchair_dump/__init__.py +17 -0
- msprobe/pytorch/torchair_dump/torchair_dump.py +114 -0
- msprobe/scripts/atb/config_example.json +10 -0
- msprobe/scripts/atb/load_atb_probe.sh +101 -0
- msprobe/scripts/atb/unload_atb_probe.sh +27 -0
- msprobe/scripts/build_msaccucmp.sh +186 -0
- msprobe/scripts/conf/help.info +6 -0
- msprobe/scripts/conf/version.info +3 -0
- msprobe/scripts/run_script/common.sh +538 -0
- msprobe/scripts/run_script/main_msaccucmp.sh +232 -0
- msprobe/visualization/__init__.py +12 -11
- msprobe/visualization/builder/__init__.py +12 -11
- msprobe/visualization/builder/graph_builder.py +45 -30
- msprobe/visualization/builder/graph_merger.py +53 -32
- msprobe/visualization/builder/msprobe_adapter.py +34 -44
- msprobe/visualization/compare/__init__.py +12 -11
- msprobe/visualization/compare/graph_comparator.py +63 -51
- msprobe/visualization/compare/mode_adapter.py +28 -113
- msprobe/visualization/db_utils.py +133 -22
- msprobe/visualization/graph/__init__.py +12 -11
- msprobe/visualization/graph/base_node.py +15 -27
- msprobe/visualization/graph/distributed_analyzer.py +97 -40
- msprobe/visualization/graph/graph.py +14 -16
- msprobe/visualization/graph/node_colors.py +34 -31
- msprobe/visualization/graph/node_op.py +12 -11
- msprobe/visualization/graph_service.py +580 -205
- msprobe/visualization/utils.py +278 -31
- tb_graph_ascend/secure_build.py +175 -0
- tb_graph_ascend/server/__init__.py +15 -0
- tb_graph_ascend/server/app/__init__.py +15 -0
- tb_graph_ascend/server/app/model/__init__.py +15 -0
- tb_graph_ascend/server/app/model/hierarchy.py +348 -0
- tb_graph_ascend/server/app/model/layout_hierarchy_model.py +69 -0
- tb_graph_ascend/server/app/model/match_nodes_model.py +573 -0
- tb_graph_ascend/server/app/repositories/__init__.py +15 -0
- tb_graph_ascend/server/app/repositories/graph_repo_base.py +32 -0
- tb_graph_ascend/server/app/repositories/graph_repo_db.py +879 -0
- tb_graph_ascend/server/app/repositories/graph_repo_vis.py +83 -0
- tb_graph_ascend/server/app/service/__init__.py +18 -0
- tb_graph_ascend/server/app/service/graph_service_base.py +158 -0
- tb_graph_ascend/server/app/service/graph_service_db.py +438 -0
- tb_graph_ascend/server/app/service/graph_service_factory.py +54 -0
- tb_graph_ascend/server/app/service/graph_service_vis.py +480 -0
- tb_graph_ascend/server/app/utils/__init__.py +15 -0
- tb_graph_ascend/server/app/utils/constant.py +80 -0
- tb_graph_ascend/server/app/utils/file_check_wrapper.py +46 -0
- tb_graph_ascend/server/app/utils/global_state.py +95 -0
- tb_graph_ascend/server/app/utils/graph_utils.py +661 -0
- tb_graph_ascend/server/app/utils/i18n.py +153 -0
- tb_graph_ascend/server/app/utils/request_method.py +46 -0
- tb_graph_ascend/server/app/views/__init__.py +15 -0
- tb_graph_ascend/server/app/views/graph_views.py +304 -0
- tb_graph_ascend/server/plugin.py +108 -0
- tb_graph_ascend/server/static/index.html +9250 -0
- tb_graph_ascend/server/static/index.js +21 -0
- tb_graph_ascend/setup.py +57 -0
- mindstudio_probe-8.3.3.dist-info/LICENSE +0 -201
- mindstudio_probe-8.3.3.dist-info/RECORD +0 -491
- mindstudio_probe-8.3.3.dist-info/entry_points.txt +0 -2
- mindstudio_probe-8.3.3.dist-info/top_level.txt +0 -1
- msprobe/CMakeLists.txt +0 -5
- msprobe/README.md +0 -203
- msprobe/core/advisor/advisor.py +0 -129
- msprobe/core/advisor/advisor_const.py +0 -58
- msprobe/core/advisor/advisor_result.py +0 -58
- msprobe/core/compare/find_first/data_processor.py +0 -35
- msprobe/core/compare/highlight.py +0 -390
- msprobe/core/data_dump/data_collector.py +0 -356
- msprobe/core/grad_probe/constant.py +0 -90
- msprobe/core/grad_probe/grad_compare.py +0 -187
- msprobe/core/grad_probe/utils.py +0 -105
- msprobe/core/kernel_dump/kernel_config.py +0 -33
- msprobe/docs/01.installation.md +0 -250
- msprobe/docs/02.config_introduction.md +0 -221
- msprobe/docs/03.config_examples.md +0 -281
- msprobe/docs/04.kernel_dump_PyTorch.md +0 -73
- msprobe/docs/05.data_dump_PyTorch.md +0 -518
- msprobe/docs/06.data_dump_MindSpore.md +0 -618
- msprobe/docs/07.accuracy_checker_PyTorch.md +0 -310
- msprobe/docs/09.accuracy_checker_MindSpore.md +0 -120
- msprobe/docs/10.accuracy_compare_PyTorch.md +0 -637
- msprobe/docs/11.accuracy_compare_MindSpore.md +0 -769
- msprobe/docs/12.overflow_check_PyTorch.md +0 -82
- msprobe/docs/13.overflow_check_MindSpore.md +0 -33
- msprobe/docs/14.data_parse_PyTorch.md +0 -282
- msprobe/docs/15.free_benchmarking_PyTorch.md +0 -169
- msprobe/docs/16.free_benchmarking_MindSpore.md +0 -159
- msprobe/docs/17.grad_probe.md +0 -205
- msprobe/docs/18.online_dispatch.md +0 -89
- msprobe/docs/19.monitor.md +0 -753
- msprobe/docs/20.monitor_performance_baseline.md +0 -52
- msprobe/docs/21.visualization_PyTorch.md +0 -519
- msprobe/docs/22.visualization_MindSpore.md +0 -515
- msprobe/docs/23.generate_operator_PyTorch.md +0 -107
- msprobe/docs/24.code_mapping_Mindspore.md +0 -29
- msprobe/docs/25.tool_function_introduction.md +0 -29
- msprobe/docs/26.data_dump_PyTorch_baseline.md +0 -48
- msprobe/docs/27.dump_json_instruction.md +0 -795
- msprobe/docs/28.debugger_save_instruction.md +0 -288
- msprobe/docs/28.kernel_dump_MindSpore.md +0 -69
- msprobe/docs/29.data_dump_MSAdapter.md +0 -235
- msprobe/docs/30.overflow_check_MSAdapter.md +0 -31
- msprobe/docs/31.config_check.md +0 -107
- msprobe/docs/32.ckpt_compare.md +0 -69
- msprobe/docs/33.generate_operator_MindSpore.md +0 -181
- msprobe/docs/34.RL_collect.md +0 -101
- msprobe/docs/35.nan_analyze.md +0 -73
- msprobe/docs/36.calculation_result_change.md +0 -75
- msprobe/docs/FAQ.md +0 -232
- msprobe/docs/S02.report_free_benchmarking_validation_performance_baseline.md +0 -146
- msprobe/docs/accuracy_checker_MindSpore/accuracy_checker_MindSpore_baseline.md +0 -14
- msprobe/docs/data_dump_MindSpore/data_dump_MindSpore_baseline.md +0 -33
- msprobe/docs/data_dump_MindSpore/dynamic_graph_quick_start_example.md +0 -217
- msprobe/docs/img/BLOOM-7B_1.png +0 -0
- msprobe/docs/img/BLOOM-7B_2.png +0 -0
- msprobe/docs/img/BLOOM-7B_3.png +0 -0
- msprobe/docs/img/BLOOM-7B_4.png +0 -0
- msprobe/docs/img/GPT-3_1.png +0 -0
- msprobe/docs/img/GPT-3_2.png +0 -0
- msprobe/docs/img/GPT-3_3.png +0 -0
- msprobe/docs/img/GPT-3_4.png +0 -0
- msprobe/docs/img/GPT-3_5.png +0 -0
- msprobe/docs/img/GPT-3_6.png +0 -0
- msprobe/docs/img/GPT-3_7.png +0 -0
- msprobe/docs/img/GPT-3_8.png +0 -0
- msprobe/docs/img/YOLOV5S_1.png +0 -0
- msprobe/docs/img/YOLOV5S_2.png +0 -0
- msprobe/docs/img/accuracy_checking_details.png +0 -0
- msprobe/docs/img/accuracy_checking_result.png +0 -0
- msprobe/docs/img/api_precision_compare_details.png +0 -0
- msprobe/docs/img/api_precision_compare_result.png +0 -0
- msprobe/docs/img/auto_analyze_log.png +0 -0
- msprobe/docs/img/compare_result.png +0 -0
- msprobe/docs/img/compare_result_pkl.png +0 -0
- msprobe/docs/img/compare_result_pkl_md5.png.png +0 -0
- msprobe/docs/img/cpu_info.png +0 -0
- msprobe/docs/img/free_benchmark.png +0 -0
- msprobe/docs/img/free_benchmark_framework.png +0 -0
- msprobe/docs/img/grad_probe_image-1.png +0 -0
- msprobe/docs/img/grad_probe_image-2.png +0 -0
- msprobe/docs/img/grad_probe_image-3.png +0 -0
- msprobe/docs/img/grad_probe_image-4.png +0 -0
- msprobe/docs/img/grad_probe_image.png +0 -0
- msprobe/docs/img/merge_result.png +0 -0
- msprobe/docs/img/module_compare.png +0 -0
- msprobe/docs/img/monitor/cpu_info.png +0 -0
- msprobe/docs/img/monitor/step_count_per_record.png +0 -0
- msprobe/docs/img/ms_dump.png +0 -0
- msprobe/docs/img/ms_layer.png +0 -0
- msprobe/docs/img/pt_dump.png +0 -0
- msprobe/docs/img/save_compare_result_sample.png +0 -0
- msprobe/docs/img/visualization/fuzzy_match_ms.png +0 -0
- msprobe/docs/img/visualization/fuzzy_match_pt.png +0 -0
- msprobe/docs/img/visualization/proxy.png +0 -0
- msprobe/docs/img/visualization/tensorboard_1.png +0 -0
- msprobe/docs/img/visualization/tensorboard_2.png +0 -0
- msprobe/docs/img/visualization/vis_browser_1.png +0 -0
- msprobe/docs/img/visualization/vis_browser_2.png +0 -0
- msprobe/docs/img/visualization/vis_match_info.png +0 -0
- msprobe/docs/img/visualization/vis_precision_info.png +0 -0
- msprobe/docs/img/visualization/vis_search_info.png +0 -0
- msprobe/docs/img/visualization/vis_show_info.png +0 -0
- msprobe/docs/img/visualization/vis_showcase.png +0 -0
- msprobe/docs/img/visualization/vis_unmatch_info.png +0 -0
- msprobe/docs/visualization/GPTModel.png +0 -0
- msprobe/docs/visualization/ParallelMLP.png +0 -0
- msprobe/docs/visualization/layer_mapping_example.md +0 -132
- msprobe/docs/visualization/mapping.png +0 -0
- msprobe/docs/visualization/mapping1.png +0 -0
- msprobe/docs/visualization/mindspeed_llamafactoary_img/1.png +0 -0
- msprobe/docs/visualization/mindspeed_llamafactoary_img/2.png +0 -0
- msprobe/docs/visualization/mindspeed_llamafactoary_img/3.png +0 -0
- msprobe/docs/visualization/mindspeed_llamafactoary_img/4.png +0 -0
- msprobe/docs/visualization/mindspeed_llamafactoary_img/5.png +0 -0
- msprobe/docs/visualization/mindspeed_llamafactoary_img/6.png +0 -0
- msprobe/docs/visualization/mindspeed_llamafactoary_img/7.png +0 -0
- msprobe/docs/visualization/mindspeed_llamafactoary_img/llamafactory-qwen25vl.txt +0 -59
- msprobe/docs/visualization/mindspeed_llamafactoary_img/llamafactory1.png +0 -0
- msprobe/docs/visualization/mindspeed_llamafactoary_img/llamafactory2.png +0 -0
- msprobe/docs/visualization/mindspeed_llamafactoary_img/mindspeed-mm-qwen25vl.txt +0 -80
- msprobe/docs/visualization/mindspeed_llamafactoary_img/mindspeed1.png +0 -0
- msprobe/docs/visualization/mindspeed_llamafactoary_img/mindspeed2.png +0 -0
- msprobe/docs/visualization/mindspeed_llamafactory_mapping.md +0 -330
- msprobe/docs/visualization/module_name.png +0 -0
- msprobe/docs/visualization/module_name1.png +0 -0
- msprobe/docs/visualization/no_mapping.png +0 -0
- msprobe/docs/visualization/no_mapping1.png +0 -0
- msprobe/docs/visualization/no_mapping_analyze.png +0 -0
- msprobe/docs/visualization/top_layer.png +0 -0
- msprobe/mindspore/api_accuracy_checker/generate_op_script/op_generator.py +0 -460
- msprobe/mindspore/api_accuracy_checker/generate_op_script/operator_replication.template +0 -2081
- msprobe/mindspore/code_mapping/bind.py +0 -283
- msprobe/mindspore/code_mapping/cmd_parser.py +0 -40
- msprobe/mindspore/code_mapping/graph.py +0 -49
- msprobe/mindspore/code_mapping/graph_parser.py +0 -211
- msprobe/mindspore/code_mapping/main.py +0 -24
- msprobe/mindspore/code_mapping/processor.py +0 -34
- msprobe/mindspore/dym_loader/hook_dynamic_loader.cpp +0 -111
- msprobe/mindspore/dym_loader/hook_dynamic_loader.h +0 -52
- msprobe/mindspore/free_benchmark/api_pynative_self_check.py +0 -257
- msprobe/mindspore/free_benchmark/common/config.py +0 -27
- msprobe/mindspore/free_benchmark/common/handler_params.py +0 -31
- msprobe/mindspore/free_benchmark/common/utils.py +0 -100
- msprobe/mindspore/free_benchmark/data/support_wrap_ops.yaml +0 -638
- msprobe/mindspore/free_benchmark/handler/base_handler.py +0 -105
- msprobe/mindspore/free_benchmark/handler/check_handler.py +0 -55
- msprobe/mindspore/free_benchmark/handler/fix_handler.py +0 -51
- msprobe/mindspore/free_benchmark/handler/handler_factory.py +0 -36
- msprobe/mindspore/free_benchmark/perturbation/add_noise.py +0 -82
- msprobe/mindspore/free_benchmark/perturbation/base_perturbation.py +0 -45
- msprobe/mindspore/free_benchmark/perturbation/bit_noise.py +0 -78
- msprobe/mindspore/free_benchmark/perturbation/exchange_value.py +0 -77
- msprobe/mindspore/free_benchmark/perturbation/improve_precision.py +0 -56
- msprobe/mindspore/free_benchmark/perturbation/no_change.py +0 -27
- msprobe/mindspore/free_benchmark/perturbation/perturbation_factory.py +0 -46
- msprobe/mindspore/free_benchmark/self_check_tool_factory.py +0 -51
- msprobe/mindspore/grad_probe/global_context.py +0 -127
- msprobe/mindspore/grad_probe/grad_analyzer.py +0 -260
- msprobe/mindspore/grad_probe/grad_monitor.py +0 -42
- msprobe/mindspore/grad_probe/grad_stat_csv.py +0 -161
- msprobe/mindspore/grad_probe/hook.py +0 -115
- msprobe/mindspore/grad_probe/utils.py +0 -43
- msprobe/mindspore/mindtorch/__init__.py +0 -18
- msprobe/mindspore/ms_config.py +0 -153
- msprobe/mindspore/task_handler_factory.py +0 -44
- msprobe/nan_analyze/__init__.py +0 -14
- msprobe/pytorch/api_accuracy_checker/generate_op_script/config_op.json +0 -9
- msprobe/pytorch/api_accuracy_checker/generate_op_script/op_generator.py +0 -480
- msprobe/pytorch/api_accuracy_checker/generate_op_script/operator_replication.template +0 -567
- msprobe/pytorch/debugger/precision_debugger.py +0 -181
- msprobe/pytorch/free_benchmark/__init__.py +0 -23
- msprobe/pytorch/free_benchmark/common/constant.py +0 -85
- msprobe/pytorch/free_benchmark/common/counter.py +0 -87
- msprobe/pytorch/free_benchmark/common/enums.py +0 -80
- msprobe/pytorch/free_benchmark/common/params.py +0 -152
- msprobe/pytorch/free_benchmark/common/utils.py +0 -143
- msprobe/pytorch/free_benchmark/compare/grad_saver.py +0 -215
- msprobe/pytorch/free_benchmark/compare/single_benchmark.py +0 -121
- msprobe/pytorch/free_benchmark/main.py +0 -123
- msprobe/pytorch/free_benchmark/perturbed_layers/base_layer.py +0 -28
- msprobe/pytorch/free_benchmark/perturbed_layers/layer_factory.py +0 -56
- msprobe/pytorch/free_benchmark/perturbed_layers/npu/add_noise.py +0 -107
- msprobe/pytorch/free_benchmark/perturbed_layers/npu/bit_noise.py +0 -121
- msprobe/pytorch/free_benchmark/perturbed_layers/npu/change_value.py +0 -89
- msprobe/pytorch/free_benchmark/perturbed_layers/npu/improve_precision.py +0 -87
- msprobe/pytorch/free_benchmark/perturbed_layers/npu/no_change.py +0 -43
- msprobe/pytorch/free_benchmark/perturbed_layers/npu/npu_base_layser.py +0 -60
- msprobe/pytorch/free_benchmark/perturbed_layers/run_cpu.py +0 -34
- msprobe/pytorch/free_benchmark/result_handlers/base_handler.py +0 -252
- msprobe/pytorch/free_benchmark/result_handlers/check_handler.py +0 -54
- msprobe/pytorch/free_benchmark/result_handlers/fix_handler.py +0 -40
- msprobe/pytorch/free_benchmark/result_handlers/handler_factory.py +0 -45
- msprobe/pytorch/free_benchmark/result_handlers/preheat_handler.py +0 -181
- msprobe/pytorch/grad_probe/__init__.py +0 -0
- msprobe/pytorch/grad_probe/grad_monitor.py +0 -108
- msprobe/pytorch/grad_probe/grad_stat_csv.py +0 -160
- msprobe/pytorch/hook_module/__init__.py +0 -16
- msprobe/pytorch/hook_module/wrap_aten.py +0 -111
- msprobe/pytorch/online_dispatch/__init__.py +0 -19
- msprobe/pytorch/online_dispatch/compare.py +0 -224
- msprobe/pytorch/online_dispatch/dispatch.py +0 -332
- msprobe/pytorch/online_dispatch/dump_compare.py +0 -179
- msprobe/pytorch/online_dispatch/single_compare.py +0 -412
- msprobe/pytorch/online_dispatch/torch_ops_config.yaml +0 -58
- msprobe/pytorch/online_dispatch/utils.py +0 -158
- msprobe/pytorch/parse_tool/__init__.py +0 -0
- msprobe/pytorch/parse_tool/cli.py +0 -31
- msprobe/pytorch/parse_tool/lib/__init__.py +0 -0
- msprobe/pytorch/parse_tool/lib/compare.py +0 -253
- msprobe/pytorch/parse_tool/lib/config.py +0 -50
- msprobe/pytorch/parse_tool/lib/file_desc.py +0 -45
- msprobe/pytorch/parse_tool/lib/interactive_cli.py +0 -97
- msprobe/pytorch/parse_tool/lib/parse_exception.py +0 -54
- msprobe/pytorch/parse_tool/lib/parse_tool.py +0 -161
- msprobe/pytorch/parse_tool/lib/utils.py +0 -299
- msprobe/pytorch/parse_tool/lib/visualization.py +0 -85
- msprobe/pytorch/pt_config.py +0 -299
- /msprobe/core/{grad_probe → dump}/__init__.py +0 -0
- /msprobe/{mindspore/code_mapping → core/dump/api_dump}/__init__.py +0 -0
- /msprobe/{mindspore/debugger → core/dump/data_dump}/__init__.py +0 -0
- /msprobe/{mindspore/exception_dump → core/dump/data_dump/data_processor}/__init__.py +0 -0
- /msprobe/{mindspore/free_benchmark → core/dump/debugger}/__init__.py +0 -0
- /msprobe/{mindspore/free_benchmark/common → core/dump/kernel_dump}/__init__.py +0 -0
- /msprobe/mindspore/{free_benchmark/handler → dump/debugger}/__init__.py +0 -0
- /msprobe/mindspore/{grad_probe → dump/dump_processor}/__init__.py +0 -0
- /msprobe/mindspore/{overflow_check → dump/exception_dump}/__init__.py +0 -0
- /msprobe/mindspore/{mindtorch → dump/mindtorch}/mindtorch_adaptor.py +0 -0
- /msprobe/{pytorch/api_accuracy_checker/run_ut → mindspore/dump/overflow_check}/__init__.py +0 -0
- /msprobe/{pytorch/debugger → mindspore/monitor}/__init__.py +0 -0
- /msprobe/{pytorch/free_benchmark/common → msaccucmp}/__init__.py +0 -0
- /msprobe/pytorch/api_accuracy_checker/{run_ut → acc_check}/.keep +0 -0
- /msprobe/pytorch/{free_benchmark/perturbed_layers → api_accuracy_checker/acc_check}/__init__.py +0 -0
- /msprobe/pytorch/api_accuracy_checker/{run_ut → acc_check}/torch_ut_setting.json +0 -0
- /msprobe/pytorch/{free_benchmark/perturbed_layers/npu → dump/api_dump}/__init__.py +0 -0
- /msprobe/pytorch/{hook_module → dump/api_dump}/support_wrap_ops.yaml +0 -0
- /msprobe/pytorch/{free_benchmark/result_handlers → dump/debugger}/__init__.py +0 -0
|
@@ -1,17 +1,18 @@
|
|
|
1
|
-
#
|
|
2
|
-
#
|
|
1
|
+
# -------------------------------------------------------------------------
|
|
2
|
+
# This file is part of the MindStudio project.
|
|
3
|
+
# Copyright (c) 2025 Huawei Technologies Co.,Ltd.
|
|
3
4
|
#
|
|
4
|
-
#
|
|
5
|
-
#
|
|
6
|
-
# You may obtain a copy of
|
|
5
|
+
# MindStudio is licensed under Mulan PSL v2.
|
|
6
|
+
# You can use this software according to the terms and conditions of the Mulan PSL v2.
|
|
7
|
+
# You may obtain a copy of Mulan PSL v2 at:
|
|
7
8
|
#
|
|
8
|
-
#
|
|
9
|
+
# http://license.coscl.org.cn/MulanPSL2
|
|
9
10
|
#
|
|
10
|
-
#
|
|
11
|
-
#
|
|
12
|
-
#
|
|
13
|
-
# See the
|
|
14
|
-
#
|
|
11
|
+
# THIS SOFTWARE IS PROVIDED ON AN "AS IS" BASIS, WITHOUT WARRANTIES OF ANY KIND,
|
|
12
|
+
# EITHER EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO NON-INFRINGEMENT,
|
|
13
|
+
# MERCHANTABILITY OR FIT FOR A PARTICULAR PURPOSE.
|
|
14
|
+
# See the Mulan PSL v2 for more details.
|
|
15
|
+
# -------------------------------------------------------------------------
|
|
15
16
|
|
|
16
17
|
dependency:
|
|
17
18
|
- transformers
|
|
@@ -19,4 +20,6 @@ dependency:
|
|
|
19
20
|
- megatron
|
|
20
21
|
- numpy
|
|
21
22
|
- datasets
|
|
22
|
-
- peft
|
|
23
|
+
- peft
|
|
24
|
+
- mindspore
|
|
25
|
+
- torch
|
|
@@ -1,17 +1,18 @@
|
|
|
1
|
-
#
|
|
2
|
-
#
|
|
1
|
+
# -------------------------------------------------------------------------
|
|
2
|
+
# This file is part of the MindStudio project.
|
|
3
|
+
# Copyright (c) 2025 Huawei Technologies Co.,Ltd.
|
|
3
4
|
#
|
|
4
|
-
#
|
|
5
|
-
#
|
|
6
|
-
# You may obtain a copy of
|
|
5
|
+
# MindStudio is licensed under Mulan PSL v2.
|
|
6
|
+
# You can use this software according to the terms and conditions of the Mulan PSL v2.
|
|
7
|
+
# You may obtain a copy of Mulan PSL v2 at:
|
|
7
8
|
#
|
|
8
|
-
#
|
|
9
|
+
# http://license.coscl.org.cn/MulanPSL2
|
|
9
10
|
#
|
|
10
|
-
#
|
|
11
|
-
#
|
|
12
|
-
#
|
|
13
|
-
# See the
|
|
14
|
-
#
|
|
11
|
+
# THIS SOFTWARE IS PROVIDED ON AN "AS IS" BASIS, WITHOUT WARRANTIES OF ANY KIND,
|
|
12
|
+
# EITHER EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO NON-INFRINGEMENT,
|
|
13
|
+
# MERCHANTABILITY OR FIT FOR A PARTICULAR PURPOSE.
|
|
14
|
+
# See the Mulan PSL v2 for more details.
|
|
15
|
+
# -------------------------------------------------------------------------
|
|
15
16
|
|
|
16
17
|
HCCL_DETERMINISTIC:
|
|
17
18
|
npu:
|
|
@@ -1,17 +1,18 @@
|
|
|
1
|
-
#
|
|
2
|
-
#
|
|
1
|
+
# -------------------------------------------------------------------------
|
|
2
|
+
# This file is part of the MindStudio project.
|
|
3
|
+
# Copyright (c) 2025 Huawei Technologies Co.,Ltd.
|
|
3
4
|
#
|
|
4
|
-
#
|
|
5
|
-
#
|
|
6
|
-
# You may obtain a copy of
|
|
5
|
+
# MindStudio is licensed under Mulan PSL v2.
|
|
6
|
+
# You can use this software according to the terms and conditions of the Mulan PSL v2.
|
|
7
|
+
# You may obtain a copy of Mulan PSL v2 at:
|
|
7
8
|
#
|
|
8
|
-
#
|
|
9
|
+
# http://license.coscl.org.cn/MulanPSL2
|
|
9
10
|
#
|
|
10
|
-
#
|
|
11
|
-
#
|
|
12
|
-
#
|
|
13
|
-
# See the
|
|
14
|
-
#
|
|
11
|
+
# THIS SOFTWARE IS PROVIDED ON AN "AS IS" BASIS, WITHOUT WARRANTIES OF ANY KIND,
|
|
12
|
+
# EITHER EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO NON-INFRINGEMENT,
|
|
13
|
+
# MERCHANTABILITY OR FIT FOR A PARTICULAR PURPOSE.
|
|
14
|
+
# See the Mulan PSL v2 for more details.
|
|
15
|
+
# -------------------------------------------------------------------------
|
|
15
16
|
|
|
16
17
|
import re
|
|
17
18
|
from abc import ABC, abstractmethod
|
|
@@ -1,17 +1,18 @@
|
|
|
1
|
-
#
|
|
2
|
-
#
|
|
1
|
+
# -------------------------------------------------------------------------
|
|
2
|
+
# This file is part of the MindStudio project.
|
|
3
|
+
# Copyright (c) 2025 Huawei Technologies Co.,Ltd.
|
|
3
4
|
#
|
|
4
|
-
#
|
|
5
|
-
#
|
|
6
|
-
# You may obtain a copy of
|
|
5
|
+
# MindStudio is licensed under Mulan PSL v2.
|
|
6
|
+
# You can use this software according to the terms and conditions of the Mulan PSL v2.
|
|
7
|
+
# You may obtain a copy of Mulan PSL v2 at:
|
|
7
8
|
#
|
|
8
|
-
#
|
|
9
|
+
# http://license.coscl.org.cn/MulanPSL2
|
|
9
10
|
#
|
|
10
|
-
#
|
|
11
|
-
#
|
|
12
|
-
#
|
|
13
|
-
# See the
|
|
14
|
-
#
|
|
11
|
+
# THIS SOFTWARE IS PROVIDED ON AN "AS IS" BASIS, WITHOUT WARRANTIES OF ANY KIND,
|
|
12
|
+
# EITHER EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO NON-INFRINGEMENT,
|
|
13
|
+
# MERCHANTABILITY OR FIT FOR A PARTICULAR PURPOSE.
|
|
14
|
+
# See the Mulan PSL v2 for more details.
|
|
15
|
+
# -------------------------------------------------------------------------
|
|
15
16
|
|
|
16
17
|
import os
|
|
17
18
|
import re
|
|
@@ -1,17 +1,18 @@
|
|
|
1
|
-
#
|
|
2
|
-
#
|
|
1
|
+
# -------------------------------------------------------------------------
|
|
2
|
+
# This file is part of the MindStudio project.
|
|
3
|
+
# Copyright (c) 2025 Huawei Technologies Co.,Ltd.
|
|
3
4
|
#
|
|
4
|
-
#
|
|
5
|
-
#
|
|
6
|
-
# You may obtain a copy of
|
|
5
|
+
# MindStudio is licensed under Mulan PSL v2.
|
|
6
|
+
# You can use this software according to the terms and conditions of the Mulan PSL v2.
|
|
7
|
+
# You may obtain a copy of Mulan PSL v2 at:
|
|
7
8
|
#
|
|
8
|
-
#
|
|
9
|
+
# http://license.coscl.org.cn/MulanPSL2
|
|
9
10
|
#
|
|
10
|
-
#
|
|
11
|
-
#
|
|
12
|
-
#
|
|
13
|
-
# See the
|
|
14
|
-
#
|
|
11
|
+
# THIS SOFTWARE IS PROVIDED ON AN "AS IS" BASIS, WITHOUT WARRANTIES OF ANY KIND,
|
|
12
|
+
# EITHER EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO NON-INFRINGEMENT,
|
|
13
|
+
# MERCHANTABILITY OR FIT FOR A PARTICULAR PURPOSE.
|
|
14
|
+
# See the Mulan PSL v2 for more details.
|
|
15
|
+
# -------------------------------------------------------------------------
|
|
15
16
|
|
|
16
17
|
import inspect
|
|
17
18
|
from typing import Dict, Any, Optional, Callable, Union, List, Tuple
|
|
@@ -1,17 +1,18 @@
|
|
|
1
|
-
#
|
|
2
|
-
#
|
|
1
|
+
# -------------------------------------------------------------------------
|
|
2
|
+
# This file is part of the MindStudio project.
|
|
3
|
+
# Copyright (c) 2025 Huawei Technologies Co.,Ltd.
|
|
3
4
|
#
|
|
4
|
-
#
|
|
5
|
-
#
|
|
6
|
-
# You may obtain a copy of
|
|
5
|
+
# MindStudio is licensed under Mulan PSL v2.
|
|
6
|
+
# You can use this software according to the terms and conditions of the Mulan PSL v2.
|
|
7
|
+
# You may obtain a copy of Mulan PSL v2 at:
|
|
7
8
|
#
|
|
8
|
-
#
|
|
9
|
+
# http://license.coscl.org.cn/MulanPSL2
|
|
9
10
|
#
|
|
10
|
-
#
|
|
11
|
-
#
|
|
12
|
-
#
|
|
13
|
-
# See the
|
|
14
|
-
#
|
|
11
|
+
# THIS SOFTWARE IS PROVIDED ON AN "AS IS" BASIS, WITHOUT WARRANTIES OF ANY KIND,
|
|
12
|
+
# EITHER EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO NON-INFRINGEMENT,
|
|
13
|
+
# MERCHANTABILITY OR FIT FOR A PARTICULAR PURPOSE.
|
|
14
|
+
# See the Mulan PSL v2 for more details.
|
|
15
|
+
# -------------------------------------------------------------------------
|
|
15
16
|
|
|
16
17
|
import re
|
|
17
18
|
|
|
@@ -28,7 +29,6 @@ class CommonConfig:
|
|
|
28
29
|
self.rank = get_real_step_or_rank(json_config.get('rank'), Const.RANK)
|
|
29
30
|
self.step = get_real_step_or_rank(json_config.get('step'), Const.STEP)
|
|
30
31
|
self.level = json_config.get('level')
|
|
31
|
-
self.enable_dataloader = json_config.get('enable_dataloader', False)
|
|
32
32
|
self.async_dump = json_config.get("async_dump", False)
|
|
33
33
|
self.precision = json_config.get("precision", Const.DUMP_PRECISION_LOW)
|
|
34
34
|
self._check_config()
|
|
@@ -43,9 +43,6 @@ class CommonConfig:
|
|
|
43
43
|
if self.level and self.level not in Const.LEVEL_LIST:
|
|
44
44
|
logger.error_log_with_exp("level is invalid, it should be one of {}".format(Const.LEVEL_LIST),
|
|
45
45
|
MsprobeException(MsprobeException.INVALID_PARAM_ERROR))
|
|
46
|
-
if not isinstance(self.enable_dataloader, bool):
|
|
47
|
-
logger.error_log_with_exp("enable_dataloader is invalid, it should be a boolean",
|
|
48
|
-
MsprobeException(MsprobeException.INVALID_PARAM_ERROR))
|
|
49
46
|
if not isinstance(self.async_dump, bool):
|
|
50
47
|
logger.error_log_with_exp("async_dump is invalid, it should be a boolean",
|
|
51
48
|
MsprobeException(MsprobeException.INVALID_PARAM_ERROR))
|
|
@@ -64,16 +61,8 @@ class BaseConfig:
|
|
|
64
61
|
self.data_mode = json_config.get('data_mode')
|
|
65
62
|
self.file_format = json_config.get("file_format")
|
|
66
63
|
self.summary_mode = json_config.get("summary_mode")
|
|
67
|
-
self.
|
|
64
|
+
self.diff_nums = json_config.get("diff_nums")
|
|
68
65
|
self.check_mode = json_config.get("check_mode")
|
|
69
|
-
self.fuzz_device = json_config.get("fuzz_device")
|
|
70
|
-
self.pert_mode = json_config.get("pert_mode")
|
|
71
|
-
self.handler_type = json_config.get("handler_type")
|
|
72
|
-
self.fuzz_level = json_config.get("fuzz_level")
|
|
73
|
-
self.fuzz_stage = json_config.get("fuzz_stage")
|
|
74
|
-
self.if_preheat = json_config.get("if_preheat")
|
|
75
|
-
self.preheat_step = json_config.get("preheat_step")
|
|
76
|
-
self.max_sample = json_config.get("max_sample")
|
|
77
66
|
self.is_regex_valid = True
|
|
78
67
|
|
|
79
68
|
@staticmethod
|
|
@@ -0,0 +1,257 @@
|
|
|
1
|
+
# -------------------------------------------------------------------------
|
|
2
|
+
# This file is part of the MindStudio project.
|
|
3
|
+
# Copyright (c) 2025 Huawei Technologies Co.,Ltd.
|
|
4
|
+
#
|
|
5
|
+
# MindStudio is licensed under Mulan PSL v2.
|
|
6
|
+
# You can use this software according to the terms and conditions of the Mulan PSL v2.
|
|
7
|
+
# You may obtain a copy of Mulan PSL v2 at:
|
|
8
|
+
#
|
|
9
|
+
# http://license.coscl.org.cn/MulanPSL2
|
|
10
|
+
#
|
|
11
|
+
# THIS SOFTWARE IS PROVIDED ON AN "AS IS" BASIS, WITHOUT WARRANTIES OF ANY KIND,
|
|
12
|
+
# EITHER EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO NON-INFRINGEMENT,
|
|
13
|
+
# MERCHANTABILITY OR FIT FOR A PARTICULAR PURPOSE.
|
|
14
|
+
# See the Mulan PSL v2 for more details.
|
|
15
|
+
# -------------------------------------------------------------------------
|
|
16
|
+
|
|
17
|
+
import atexit
|
|
18
|
+
import os
|
|
19
|
+
import threading
|
|
20
|
+
import traceback
|
|
21
|
+
|
|
22
|
+
from msprobe.core.common.const import Const
|
|
23
|
+
from msprobe.core.common.log import logger
|
|
24
|
+
from msprobe.core.common.megatron_utils import MegatronStepInfo, get_micro_step, is_megatron
|
|
25
|
+
from msprobe.core.dump.data_dump.data_processor.factory import DataProcessorFactory
|
|
26
|
+
from msprobe.core.dump.data_dump.json_writer import DataWriter
|
|
27
|
+
from msprobe.core.dump.data_dump.scope import ScopeFactory
|
|
28
|
+
|
|
29
|
+
|
|
30
|
+
def build_data_collector(config):
|
|
31
|
+
return DataCollector(config)
|
|
32
|
+
|
|
33
|
+
|
|
34
|
+
class DataCollector:
|
|
35
|
+
tasks_need_tensor_data = [Const.DIFF_CHECK, Const.TENSOR]
|
|
36
|
+
level_without_construct = [Const.LEVEL_L1, Const.LEVEL_L2]
|
|
37
|
+
|
|
38
|
+
def __init__(self, config):
|
|
39
|
+
self.config = config
|
|
40
|
+
self.data_writer = DataWriter()
|
|
41
|
+
self.data_processor = DataProcessorFactory.create_processor(self.config, self.data_writer)
|
|
42
|
+
self.module_processor = DataProcessorFactory.get_module_processor(self.config.framework)
|
|
43
|
+
self.module_count = {}
|
|
44
|
+
self.scope = ScopeFactory(self.config).build_scope()
|
|
45
|
+
self.backward_module_names = {}
|
|
46
|
+
self.params_grad_record = {}
|
|
47
|
+
self.optimizer_status = ""
|
|
48
|
+
self.optimizer_status_first_start = {Const.OPTIMIZER: True, Const.CLIP_GRAD: True}
|
|
49
|
+
atexit.register(self.write_json_at_exit)
|
|
50
|
+
|
|
51
|
+
@property
|
|
52
|
+
def dump_data_dir(self):
|
|
53
|
+
return self.data_writer.dump_tensor_data_dir
|
|
54
|
+
|
|
55
|
+
@property
|
|
56
|
+
def dump_file_path(self):
|
|
57
|
+
return self.data_writer.dump_file_path
|
|
58
|
+
|
|
59
|
+
@property
|
|
60
|
+
def bench_dump_file_path(self):
|
|
61
|
+
return self.data_writer.bench_dump_file_path
|
|
62
|
+
|
|
63
|
+
@staticmethod
|
|
64
|
+
def check_scope_and_pid(scope, name, pid):
|
|
65
|
+
return (not scope or scope.check(name)) and pid == os.getpid()
|
|
66
|
+
|
|
67
|
+
@staticmethod
|
|
68
|
+
def set_is_recomputable(data_info, is_recompute):
|
|
69
|
+
if data_info and len(data_info) == 1 and is_recompute is not None: # 正常情况下data_info的长度应改为1
|
|
70
|
+
data_info[list(data_info.keys())[0]]["is_recompute"] = is_recompute
|
|
71
|
+
|
|
72
|
+
def reset_status(self):
|
|
73
|
+
self.optimizer_status = ""
|
|
74
|
+
self.optimizer_status_first_start = {Const.OPTIMIZER: True, Const.CLIP_GRAD: True}
|
|
75
|
+
self.data_writer.reset_cache()
|
|
76
|
+
self.backward_module_names.clear()
|
|
77
|
+
|
|
78
|
+
def update_api_or_module_name(self, api_or_module_name):
|
|
79
|
+
self.data_processor.update_api_or_module_name(api_or_module_name)
|
|
80
|
+
|
|
81
|
+
def write_json(self):
|
|
82
|
+
self.data_writer.write_json()
|
|
83
|
+
|
|
84
|
+
def write_json_at_exit(self):
|
|
85
|
+
if self.config.async_dump and self.config.task == Const.TENSOR:
|
|
86
|
+
self.data_processor.dump_async_data()
|
|
87
|
+
self.data_writer.write_json()
|
|
88
|
+
|
|
89
|
+
def update_data(self, name, data_info):
|
|
90
|
+
logger.debug(f"msprobe is collecting data on {name}.")
|
|
91
|
+
self.data_writer.update_data(data_info)
|
|
92
|
+
|
|
93
|
+
def call_stack_collect(self, data_info, name):
|
|
94
|
+
stack_info, is_recompute = self.data_processor.analyze_api_call_stack(name)
|
|
95
|
+
self.data_writer.update_stack(name, stack_info)
|
|
96
|
+
self.set_is_recomputable(data_info, is_recompute)
|
|
97
|
+
|
|
98
|
+
def forward_input_data_collect(self, name, module, pid, module_input_output):
|
|
99
|
+
try:
|
|
100
|
+
if not self.check_scope_and_pid(self.scope, name, pid):
|
|
101
|
+
return
|
|
102
|
+
|
|
103
|
+
data_info = {}
|
|
104
|
+
if self.config.task != Const.STRUCTURE:
|
|
105
|
+
data_info = self.data_processor.analyze_forward_input(name, module, module_input_output)
|
|
106
|
+
if self.config.level == Const.LEVEL_L2:
|
|
107
|
+
return
|
|
108
|
+
self.call_stack_collect(data_info, name)
|
|
109
|
+
self.handle_data(name, data_info, flush=self.data_processor.is_terminated)
|
|
110
|
+
|
|
111
|
+
except Exception as e:
|
|
112
|
+
# 取异常类名作为“类型”做去重
|
|
113
|
+
error_type = type(e).__name__
|
|
114
|
+
tb = traceback.format_exc()
|
|
115
|
+
self.data_writer.write_error_log(
|
|
116
|
+
f"[ERROR] forward_input_data_collect failed: name={name}, pid={pid}\n{tb}",
|
|
117
|
+
error_type=error_type
|
|
118
|
+
)
|
|
119
|
+
|
|
120
|
+
def forward_output_data_collect(self, name, module, pid, module_input_output):
|
|
121
|
+
self.update_construct(name)
|
|
122
|
+
if not self.check_scope_and_pid(self.scope, name, pid):
|
|
123
|
+
return
|
|
124
|
+
|
|
125
|
+
data_info = {}
|
|
126
|
+
if self.config.task != Const.STRUCTURE:
|
|
127
|
+
data_info = self.data_processor.analyze_forward_output(name, module, module_input_output)
|
|
128
|
+
is_recompute = self.data_processor.is_recompute()
|
|
129
|
+
self.set_is_recomputable(data_info, is_recompute)
|
|
130
|
+
if self.config.level == Const.LEVEL_L2:
|
|
131
|
+
return
|
|
132
|
+
self.handle_data(name, data_info, flush=self.data_processor.is_terminated)
|
|
133
|
+
|
|
134
|
+
def forward_data_collect_only_tensor(self, name, module, pid, module_input_output):
|
|
135
|
+
if not self.check_scope_and_pid(self.scope, name, pid):
|
|
136
|
+
return
|
|
137
|
+
self.data_processor.analyze_forward(name, module, module_input_output)
|
|
138
|
+
|
|
139
|
+
|
|
140
|
+
def forward_data_collect(self, name, module, pid, module_input_output):
|
|
141
|
+
self.update_construct(name)
|
|
142
|
+
if not self.check_scope_and_pid(self.scope, name, pid):
|
|
143
|
+
return
|
|
144
|
+
data_info = {}
|
|
145
|
+
if self.config.task != Const.STRUCTURE:
|
|
146
|
+
data_info = self.data_processor.analyze_forward(name, module, module_input_output)
|
|
147
|
+
self.call_stack_collect(data_info, name)
|
|
148
|
+
self.handle_data(name, data_info, flush=self.data_processor.is_terminated)
|
|
149
|
+
|
|
150
|
+
|
|
151
|
+
def backward_data_collect_only_tensor(self, name, module, pid, module_input_output):
|
|
152
|
+
if not self.check_scope_and_pid(self.scope, name, pid):
|
|
153
|
+
return
|
|
154
|
+
self.data_processor.analyze_backward(name, module, module_input_output)
|
|
155
|
+
|
|
156
|
+
|
|
157
|
+
def backward_data_collect(self, name, module, pid, module_input_output):
|
|
158
|
+
self.update_construct(name)
|
|
159
|
+
if not self.check_scope_and_pid(self.scope, name, pid):
|
|
160
|
+
return
|
|
161
|
+
data_info = {}
|
|
162
|
+
if self.config.task != Const.STRUCTURE:
|
|
163
|
+
data_info = self.data_processor.analyze_backward(name, module, module_input_output)
|
|
164
|
+
if self.config.level == Const.LEVEL_L2:
|
|
165
|
+
return
|
|
166
|
+
if data_info and name.split(Const.SEP)[0] in Const.MODULE_PREFIX:
|
|
167
|
+
module_name = name.rsplit(Const.SEP, 2)[0]
|
|
168
|
+
self.backward_module_names[module_name] = True
|
|
169
|
+
self.handle_data(name, data_info, flush=self.data_processor.is_terminated)
|
|
170
|
+
|
|
171
|
+
|
|
172
|
+
def backward_input_data_collect(self, name, module, pid, module_input_output):
|
|
173
|
+
self.update_construct(name)
|
|
174
|
+
if not self.check_scope_and_pid(self.scope, name, pid):
|
|
175
|
+
return
|
|
176
|
+
data_info = {}
|
|
177
|
+
if self.config.task != Const.STRUCTURE:
|
|
178
|
+
data_info = self.data_processor.analyze_backward_input(name, module, module_input_output)
|
|
179
|
+
self.handle_data(name, data_info)
|
|
180
|
+
|
|
181
|
+
|
|
182
|
+
def backward_output_data_collect(self, name, module, pid, module_input_output):
|
|
183
|
+
self.update_construct(name)
|
|
184
|
+
if not self.check_scope_and_pid(self.scope, name, pid):
|
|
185
|
+
return
|
|
186
|
+
data_info = {}
|
|
187
|
+
if self.config.task != Const.STRUCTURE:
|
|
188
|
+
data_info = self.data_processor.analyze_backward_output(name, module, module_input_output)
|
|
189
|
+
self.handle_data(name, data_info)
|
|
190
|
+
|
|
191
|
+
|
|
192
|
+
def update_construct(self, name):
|
|
193
|
+
if self.config.level not in DataCollector.level_without_construct:
|
|
194
|
+
if self.optimizer_status in [Const.OPTIMIZER, Const.CLIP_GRAD]:
|
|
195
|
+
if self.optimizer_status_first_start[self.optimizer_status]:
|
|
196
|
+
self.data_writer.update_construct(
|
|
197
|
+
{self.optimizer_status: None if not is_megatron() else [None, get_micro_step()]})
|
|
198
|
+
self.optimizer_status_first_start[self.optimizer_status] = False
|
|
199
|
+
self.data_writer.update_construct(
|
|
200
|
+
{name: self.optimizer_status if not is_megatron() else [self.optimizer_status, get_micro_step()]})
|
|
201
|
+
else:
|
|
202
|
+
if self.config.level == Const.LEVEL_MIX and \
|
|
203
|
+
not (name.startswith(Const.MODULE) or name.startswith(Const.CELL)):
|
|
204
|
+
self.data_writer.update_construct(
|
|
205
|
+
{name: self.module_processor.api_parent_node.get(threading.get_ident())}
|
|
206
|
+
)
|
|
207
|
+
if MegatronStepInfo.is_megatron:
|
|
208
|
+
micro_step_number = max(MegatronStepInfo.forward_micro_step, MegatronStepInfo.backward_micro_step)
|
|
209
|
+
self.data_writer.update_construct({Const.MEGATRON_MICRO_STEP_NUMBER: micro_step_number})
|
|
210
|
+
|
|
211
|
+
self.data_writer.update_construct(self.module_processor.module_node)
|
|
212
|
+
|
|
213
|
+
def handle_data(self, name, data_info, flush=False):
|
|
214
|
+
if data_info:
|
|
215
|
+
self.update_data(name, data_info)
|
|
216
|
+
if self.config.async_dump:
|
|
217
|
+
return
|
|
218
|
+
if not flush:
|
|
219
|
+
self.data_writer.flush_data_periodically()
|
|
220
|
+
else:
|
|
221
|
+
self.write_json()
|
|
222
|
+
|
|
223
|
+
def update_dump_paths(self, *args):
|
|
224
|
+
self.data_writer.update_dump_paths(*args)
|
|
225
|
+
|
|
226
|
+
def initialize_json_file(self, framework=Const.UNKNOWN_FRAMEWORK):
|
|
227
|
+
self.data_writer.initialize_json_file(task=self.config.task, level=self.config.level, framework=framework)
|
|
228
|
+
|
|
229
|
+
def update_iter(self, current_iter):
|
|
230
|
+
self.data_processor.update_iter(current_iter)
|
|
231
|
+
|
|
232
|
+
def params_data_collect(self, name, param_name, pid, data, index):
|
|
233
|
+
grad_name = name + Const.SEP + Const.PARAMS_GRAD + Const.SEP + index
|
|
234
|
+
self.update_api_or_module_name(grad_name)
|
|
235
|
+
if not self.check_scope_and_pid(self.scope, name, pid) and not self.backward_module_names.get(name):
|
|
236
|
+
if self.data_writer.cache_data.get("data"):
|
|
237
|
+
self.data_writer.cache_data.get("data").pop(grad_name, None)
|
|
238
|
+
self.params_grad_record[grad_name] = False
|
|
239
|
+
return
|
|
240
|
+
data_info = self.data_processor.analyze_params(grad_name, param_name, data)
|
|
241
|
+
self.handle_data(grad_name, data_info, flush=self.data_processor.is_terminated)
|
|
242
|
+
self.params_grad_record[grad_name] = False
|
|
243
|
+
|
|
244
|
+
def debug_data_collect_forward(self, variable, name_with_count):
|
|
245
|
+
data_info = self.data_processor.analyze_debug_forward(variable, name_with_count)
|
|
246
|
+
name_with_count_category = name_with_count + Const.SEP + Const.DEBUG
|
|
247
|
+
self.data_writer.update_debug({name_with_count_category: data_info})
|
|
248
|
+
|
|
249
|
+
def debug_data_collect_backward(self, variable, grad_name_with_count):
|
|
250
|
+
# prepare all None nested data structure
|
|
251
|
+
all_none_data_info = self.data_processor.analyze_element_to_all_none(variable)
|
|
252
|
+
grad_name_with_count_category = grad_name_with_count + Const.SEP + Const.DEBUG
|
|
253
|
+
self.data_writer.update_debug({grad_name_with_count_category: all_none_data_info})
|
|
254
|
+
|
|
255
|
+
# register tensor backward hook
|
|
256
|
+
self.data_processor.analyze_debug_backward(variable, grad_name_with_count_category,
|
|
257
|
+
self.data_writer.cache_debug['data'])
|
|
@@ -1,21 +1,22 @@
|
|
|
1
|
-
#
|
|
2
|
-
#
|
|
1
|
+
# -------------------------------------------------------------------------
|
|
2
|
+
# This file is part of the MindStudio project.
|
|
3
|
+
# Copyright (c) 2025 Huawei Technologies Co.,Ltd.
|
|
3
4
|
#
|
|
4
|
-
#
|
|
5
|
-
#
|
|
6
|
-
# You may obtain a copy of
|
|
5
|
+
# MindStudio is licensed under Mulan PSL v2.
|
|
6
|
+
# You can use this software according to the terms and conditions of the Mulan PSL v2.
|
|
7
|
+
# You may obtain a copy of Mulan PSL v2 at:
|
|
7
8
|
#
|
|
8
|
-
#
|
|
9
|
+
# http://license.coscl.org.cn/MulanPSL2
|
|
9
10
|
#
|
|
10
|
-
#
|
|
11
|
-
#
|
|
12
|
-
#
|
|
13
|
-
# See the
|
|
14
|
-
#
|
|
11
|
+
# THIS SOFTWARE IS PROVIDED ON AN "AS IS" BASIS, WITHOUT WARRANTIES OF ANY KIND,
|
|
12
|
+
# EITHER EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO NON-INFRINGEMENT,
|
|
13
|
+
# MERCHANTABILITY OR FIT FOR A PARTICULAR PURPOSE.
|
|
14
|
+
# See the Mulan PSL v2 for more details.
|
|
15
|
+
# -------------------------------------------------------------------------
|
|
15
16
|
|
|
16
17
|
import copy
|
|
17
|
-
import inspect
|
|
18
18
|
import os
|
|
19
|
+
import inspect
|
|
19
20
|
from dataclasses import dataclass, is_dataclass
|
|
20
21
|
from functools import partial
|
|
21
22
|
from typing import Tuple, Dict, Optional, Any, Union
|
|
@@ -115,30 +116,6 @@ class BaseDataProcessor:
|
|
|
115
116
|
def is_terminated(self):
|
|
116
117
|
return False
|
|
117
118
|
|
|
118
|
-
@staticmethod
|
|
119
|
-
def analyze_api_call_stack(name):
|
|
120
|
-
try:
|
|
121
|
-
if name.startswith("Primitive"):
|
|
122
|
-
api_stack = inspect.stack()[4:]
|
|
123
|
-
else:
|
|
124
|
-
api_stack = inspect.stack()[5:]
|
|
125
|
-
except Exception as e:
|
|
126
|
-
logger.warning(f"The call stack of <{name}> failed to retrieve, {e}.")
|
|
127
|
-
api_stack = None
|
|
128
|
-
stack_str = []
|
|
129
|
-
if api_stack:
|
|
130
|
-
for (_, path, line, func, code, _) in api_stack:
|
|
131
|
-
if not code:
|
|
132
|
-
continue
|
|
133
|
-
if any(filter_path in path for filter_path in Const.STACK_FILTER_KEYWORDS) and \
|
|
134
|
-
Const.CALL_STACK_FLAG not in path:
|
|
135
|
-
continue
|
|
136
|
-
stack_line = f"File {path}, line {str(line)}, in {func}, \n {code[0].strip()}"
|
|
137
|
-
stack_str.append(stack_line)
|
|
138
|
-
else:
|
|
139
|
-
stack_str.append(Const.WITHOUT_CALL_STACK)
|
|
140
|
-
return tuple(stack_str)
|
|
141
|
-
|
|
142
119
|
@staticmethod
|
|
143
120
|
def transfer_type(data):
|
|
144
121
|
dtype = str(type(data))
|
|
@@ -184,6 +161,38 @@ class BaseDataProcessor:
|
|
|
184
161
|
def is_distributed_op(module):
|
|
185
162
|
return getattr(module, "op_is_distributed", False)
|
|
186
163
|
|
|
164
|
+
@staticmethod
|
|
165
|
+
def is_recompute(call_stack=None):
|
|
166
|
+
return None
|
|
167
|
+
|
|
168
|
+
@staticmethod
|
|
169
|
+
def analyze_api_call_stack(name):
|
|
170
|
+
try:
|
|
171
|
+
call_stack = inspect.stack()
|
|
172
|
+
if name.startswith("Primitive"):
|
|
173
|
+
api_stack = call_stack[4:]
|
|
174
|
+
else:
|
|
175
|
+
api_stack = call_stack[5:]
|
|
176
|
+
except Exception as e:
|
|
177
|
+
logger.warning(f"The call stack of <{name}> failed to retrieve, {e}.")
|
|
178
|
+
api_stack = None
|
|
179
|
+
call_stack = None
|
|
180
|
+
|
|
181
|
+
stack_str = []
|
|
182
|
+
if api_stack:
|
|
183
|
+
for (_, path, line, func, code, _) in api_stack:
|
|
184
|
+
if not code:
|
|
185
|
+
continue
|
|
186
|
+
if any(filter_path in path for filter_path in Const.STACK_FILTER_KEYWORDS) and \
|
|
187
|
+
Const.CALL_STACK_FLAG not in path:
|
|
188
|
+
continue
|
|
189
|
+
stack_line = f"File {path}, line {str(line)}, in {func}, \n {code[0].strip()}"
|
|
190
|
+
stack_str.append(stack_line)
|
|
191
|
+
else:
|
|
192
|
+
stack_str.append(Const.WITHOUT_CALL_STACK)
|
|
193
|
+
del call_stack
|
|
194
|
+
return tuple(stack_str), None
|
|
195
|
+
|
|
187
196
|
@staticmethod
|
|
188
197
|
def _analyze_builtin(arg):
|
|
189
198
|
single_arg = {}
|