Huawei_Technology
/
mindspore-mindinsight
Not watched
1
0
Fork 0
Code
Releases 11 Wiki evaluate Activity
Issues 0 Pull Requests 0 Datasets Model Cloudbrain HPC
You can not select more than 25 topics Topics must start with a chinese character,a letter or number, can include dashes ('-') and can be up to 35 characters long.
1738 Commits
10 Branches
14 MB
0 Download
 
 
 
 
 
 
Tree: 81113a9227
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from '81113a9227'
${ noResults }
mindspore-mindinsight/mindinsight/debugger
jiangshuqiang 81113a9227 update pull_data and delete mode watchpoint_hit from retrieve 4 years ago
..
common !967 add DebuggerSetRecommendWatchpointsError and DebuggerConditionUnavailableError 5 years ago
conditionmgr fix param check 4 years ago
proto Add error_list when hitting watchpoint and fix log by adding image_path infomation 5 years ago
stream_cache optimize tensor value connection method 4 years ago
stream_handler optimize tensor value connection method 4 years ago
stream_operator fix argument and add default node 4 years ago
README.md update links of README 5 years ago
README_CN.md update links of README 5 years ago
__init__.py add debugger module 5 years ago
debugger_cache.py add multigraph debugger support, update contion, search category, tensor summary 5 years ago
debugger_grpc_server.py update pull_data and delete mode watchpoint_hit from retrieve 4 years ago
debugger_server.py update pull_data and delete mode watchpoint_hit from retrieve 4 years ago

README.md

Debugger Introduction

查看中文

Overview

MindSpore Debugger is a debugging tool for training in Graph Mode. It can be applied to visualize and analyze the intermediate computation results of the computational graph.

In Graph Mode training, the computation results of intermediate nodes in the computational graph can not be acquired from python layer, which makes it difficult for users to do the debugging. By applying MindSpore Debugger, users can:

  • Visualize the computational graph on the UI and analyze the output of the graph node;
  • Set conditional breakpoint to monitor training exceptions (such as Nan/Inf), if the condition (Nan/Inf etc.) is met, users can track the cause of the bug when an exception occurs;
  • Visualize and analyze the change of parameters, such as weights.

Operation Process

  • Launch MindInsight in debugger mode, and set Debugger environment variables for the training;
  • At the beginning of the training, set conditional breakpoints to monitor the outputs of the graph nodes;
  • Analyze the training progress on MindInsight Debugger UI.

Debugger Instructions

For the details of how to use MindSpore Debugger, please refer to https://www.mindspore.cn/tutorial/training/en/master/advanced_use/debugger.html.

Delete task

Are you sure you want to delete this task? Once this task is deleted, it cannot be recovered.

cancel
confirm

MindInsight为MindSpore提供了简单易用的调优调试能力。在训练过程中，可以将标量、张量、图像、计算图、模型超参、训练耗时等数据记录到文件中，通过MindInsight可视化页面进行查看及分析。

SVG Text Python Vue CSV other