MindSpore Insight Commands

View Source On Gitee

View the Command Help Information

mindinsight --help

View the Version Information

mindinsight --version

Start the Service

MindSpore Insight service only supports local access by default. For remote access, please modify the configuration file mindinsight\conf\constants.py. Modify the HOST in the file to the server IP, and make sure that the startup port has been opened or the firewall has been closed.

mindinsight start [-h] [--workspace <WORKSPACE>] [--port <PORT>]
                  [--url-path-prefix <URL_PATH_PREFIX>]
                  [--reload-interval <RELOAD_INTERVAL>]
                  [--summary-base-dir <SUMMARY_BASE_DIR>]
                  [--enable-debugger <ENABLE_DEBUGGER>]
                  [--debugger-port <DEBUGGER_PORT>]
                  [--offline-debugger-mem-limit <OFFLINE_DEBUGGER_MEMORY_LIMIT>]
                  [--max-offline-debugger-session-num <MAX_OFFLINE_DEBUGGER_SESSION_NUMBER>]

Optional parameters are as follows:

Name

Argument

Description

Type

Default

Scope

Specifications

-h, --help

Optional

Displays the help information about the start command.

-

-

-

-

--workspace <WORKSPACE>

Optional

Specifies the path for storing MindSpore Insight logs.

String

$HOME/mindinsight

-

-

--port <PORT>

Optional

Specifies the port number of the web visualization service.

Integer

8080

1~65535

-

--url-path-prefix <URL_PATH_PREFIX>

Optional

Specifies the URL path prefix of the web visualization service.

String

Empty string

-

URL path prefix consists of segments separated by slashes. Each segment supports alphabets / digits / underscores / dashes / dots, but not single dot or double dots.

--reload-interval <RELOAD_INTERVAL>

Optional

Specifies the interval (unit: second) for loading data.

Integer

3

0~300

The value 0 indicates that data is loaded only once.

--summary-base-dir <SUMMARY_BASE_DIR>

Optional

Specifies the root directory for loading training log data.

String

./

-

MindSpore Insight traverses the direct subdirectories in this directory and searches for log files. If a direct subdirectory contains log files, it is identified as the log file directory. If a root directory contains log files, it is identified as the log file directory. In the ModelArts development environment, this parameter can be specified as an OBS path. Please refer to ModelArts documentation for more information.

--enable-debugger <ENABLE_DEBUGGER>

Optional

Whether to launch the MindSpore Insight Debugger.

Boolean

False

True/False/1/0

The debugger entry can be shown on MindSpore Insight UI only when MindSpore Insight Debugger is launched.

--debugger-port <DEBUGGER_PORT>

Optional

Specifies the port number of the debugger server.

Integer

50051

1~65535

-

--offline-debugger-mem-limit <OFFLINE_DEBUGGER_MEMORY_LIMIT>

Optional

Specifies the maximum memory limit of a single offline debugger session. When the offline debugger cannot be executed due to insufficient memory, set it according to the device memory.

Integer

16*1024

6*1024~The upper limit of int32

-

--max-offline-debugger-session-num <MAX_OFFLINE_DEBUGGER_SESSION_NUMBER>

Optional

Specifies the maximum session number of the offline debugger. The session number refers to the amount of training jobs that can be debugged at the same time.

Integer

2

1~2

-

--max-graph-node-size <MAX_GRAPH_NODE_SIZE >

Optional

Set the maximum number of graph nodes loaded by the debugger.

Integer

100000

1~2000000

-

--workspace log directory description:

Module name

Log directory description

Log format

datavisual

Training kanban module, it records all the logs of training Kanban module.

datavisual.<PORT>.log

debugger

Debugger module, it records all logs of the debugger module.

debugger.<PORT>.log

explainer

Explain the AI module, it records all logs that explain the data parsed by the AI module.

explainer.<PORT>.log

gunicorn

Web service module, it records all logs of the Web service module.

access.<PORT>.log
error.<PORT>.log

lineage

Traceability module, it records all logs of the traceability module.

lineage.<PORT>.log

notebook

Record all logs using MindSpore Insight in the ModelArts notebook.

notebook.<PORT>.log

optimizer

Optimizer module, it records all optimizer module logs.

optimizer.<PORT>.log

parse_summary

Summary file parsing module, it records all logs when using the summary file parsing module.

parse_summary.<PORT>.log

profiler

Performance analysis module, it records all logs of the performance analysis module.

profiler.<PORT>.log

restful_api

RESTFul API module, it records all RESTFul API interaction logs.

restful_api.<PORT>.log

scripts

Start and stop the MindSpore Insight module, it records all MindSpore Insight starts and stops.

start.<PORT>.log
stop.<PORT>.log

utils

Public module, it records all logs of the public module.

utils.<PORT>.log

Note: There is one log file for each module, but when a log file exceeds 50M, it will be renamed and archived in the format of <module name>_<PORT>.log.<id> ,module name indicates the module name, PORT indicates the PORT number, and ID indicates the number of file renaming and archiving times.

When the service is started, the parameter values of the command line are saved as the environment variables of the process and start with MINDINSIGHT_, for example, MINDINSIGHT_PORT, MINDINSIGHT_WORKSPACE, etc.

Execute command:

mindinsight start --port 8000 --workspace /path/to/workspace/dir --summary-base-dir /path/to/summary/base/dir

The startup is successful if it prompts as follows:

Web address: http://127.0.0.1:8000
service start state: success

View the Service Process Information

MindSpore Insight provides user with web services. Run the following command to view the running web service process:

ps -ef | grep mindinsight

Run the following command to access the working directory WORKSPACE corresponding to the service process based on the service process ID:

lsof -p <PID> | grep access

Output the working directory WORKSPACE as follows:

gunicorn  <PID>  <USER>  <FD>  <TYPE>  <DEVICE>  <SIZE/OFF>  <NODE>  <WORKSPACE>/log/gunicorn/access.log

Stop the Service

mindinsight stop [-h] [--port PORT]

Optional parameters are as follows:

Name

Argument

Description

Type

Default

Scope

Specifications

-h, --help

Optional

Displays the help information about the stop command.

-

-

-

-

--port <PORT>

Optional

Specifies the port number of the web visualization service.

Integer

8080

1~65535

-

Execute command:

mindinsight stop --port 8000

The shutdown is successful if it prompts as follows:

Stop mindinsight service successfully

Parse Summary

MindSpore Insight provides tools for parsing summary log files. Users can save the scalars in the summary log file into a csv file and the images into a png file through the commands, which is convenient for viewing and further processing.

mindinsight parse_summary [--summary-dir] [--output]

Optional parameters are as follows:

Name

Argument

Description

Type

Default

Scope

Specifications

--summary-dir

Optional

Specifies the root directory of summary files. If the directory contains multiple summary files, only the latest summary file is parsed.

String

./

-

The summary file directory needs to be readable and executable, and the summary file needs to be readable.

--output

Optional

Specifies the root directory for saving output files.

String

./

-

-

Execute command:

mindinsight parse_summary --summary-dir ./ --output ./

The output directory structure is as follows:

└─output_{datetime}
    ├─image
    │   └─{tag}_{step}.png
    │
    └─scalar.csv

In which,

  • output_{datetime} is the output directory. The rule is ‘output_yyyyMMdd_HHmmss_SSSSSS’ including year, month, day, hour, minute, second and microseconds.

  • {tag}_{step}.png is the image in training process. ‘tag’ and ‘step’ are the tag and step in the training (special characters in tag are deleted and ‘/’ is replaced by ‘_’).

  • scalar.csv is the file which save scalars (encoding: ‘utf-8’).

Use Mindoptimizer to Tune Hyperparameters

MindSpore Insight provides parameters tuning command. The command-line interface (CLI) provides the following commands:

usage: mindoptimizer [-h] [--version] [--config <CONFIG>]
                     [--iter <ITER>]

Optional parameters are as follows:

Name

Argument

Description

Type

Default

Scope

Specifications

-h, --help

Optional

Displays the help information about the start command.

-

-

-

-

--config <CONFIG>

Required

Specifies the configuration file.

String

-

-

Physical file path (file:/path/to/config.yaml), and the file format is yaml.

--iter <ITER>

Optional

Specifies the run times for tuning parameters

Integer

1

Positive integer

-