ibm-ai-openscale-cli 0.2.250

ibm ai openscale cli

IBM Watson OpenScale"快速路径"配置工具。此工具允许用户快速开始使用Watson OpenScale。

• 如果需要，自动为ibm watson openscale提供lite计划实例
• 如果需要，自动为IBM Watson机器学习提供Lite计划实例
• 删除并重新创建ibm watson openscale datamart实例和datamart数据库模式
• （可选）将示例机器学习模型部署到wml实例
• 将示例模型实例配置为openscale，包括有效负载日志记录、公平性检查、反馈、质量检查和可解释性
• 可选地，为样本模型存储长达7天的历史有效负载、公平性和质量数据
• 上载新的反馈数据，生成100个新的实时评分预测，运行公平性和质量检查，并生成一个解释

此版本的新增功能

• 支持添加到内置示例GermanCreditisk模型中的漂移监测，包括7天的漂移监测历史记录
• 支持自定义模型配置中的漂移监测
• 其他错误修复和稳定性改进

开始之前

• 您需要一个IBM Cloud帐户。
• 创建IBM Cloud API密钥
• 如果您已经有了一个Watson机器学习（WML）实例，请确保它启用了RC，请在迁移说明中详细了解此操作。

安装

要安装，请使用pip或easy_install：

pip install -U ibm-ai-openscale-cli

或

easy_install -U ibm-ai-openscale-cli

用法

ibm-ai-openscale-cli --help

usage: ibm-ai-openscale-cli [-h] (-a APIKEY | -i IAM_TOKEN)
                            [--env {ypprod,ypqa,ys1dev,icp}]
                            [--resource-group RESOURCE_GROUP]
                            [--postgres POSTGRES] [--icd ICD] [--db2 DB2]
                            [--wml WML] [--azure-studio AZURE_STUDIO]
                            [--azure-service AZURE_SERVICE] [--spss SPSS]
                            [--custom CUSTOM] [--aws AWS]
                            [--deployment-name DEPLOYMENT_NAME]
                            [--keep-schema] [--username USERNAME]
                            [--password PASSWORD] [--url URL]
                            [--datamart-name DATAMART_NAME]
                            [--history HISTORY] [--history-only]
                            [--history-first-day HISTORY_FIRST_DAY]
                            [--model MODEL] [--list-models]
                            [--custom-model CUSTOM_MODEL]
                            [--custom-model-directory CUSTOM_MODEL_DIRECTORY]
                            [--extend] [--protect-datamart]
                            [--reset {metrics,monitors,datamart,model,all}]
                            [--verbose] [--version]
                            [--wml-plan {lite,standard,professional}]
                            [--openscale-plan {lite,standard}]

IBM Watson Openscale "express path" configuration tool. This tool allows the
user to get started quickly with Watson OpenScale: 1) If needed, provision a
Lite plan instance for IBM Watson OpenScale 2) If needed, provision a Lite
plan instance for IBM Watson Machine Learning 3) Drop and re-create the IBM
Watson OpenScale datamart instance and datamart database schema 4) Optionally,
deploy a sample machine learning model to the WML instance 5) Configure the
sample model instance to OpenScale, including payload logging, fairness
checking, feedback, quality checking, and explainability 6) Optionally, store
up to 7 days of historical payload, fairness, and quality data for the sample
model 7) Upload new feedback data, generate 100 new live scoring predictions,
run fairness and quality checks, and generate one explanation

optional arguments:
  -h, --help            show this help message and exit
  --env {ypprod,ypqa,ys1dev,icp}
                        Environment. Default "ypprod"
  --resource-group RESOURCE_GROUP
                        Resource Group to use. If not specified, then
                        "default" group is used
  --postgres POSTGRES   Path to postgres credentials file for the datamart
                        database. If --postgres, --icd, and --db2 all are not
                        specified, then the internal Watson OpenScale database
                        is used
  --icd ICD             Path to IBM Cloud Database credentials file for the
                        datamart database
  --db2 DB2             Path to IBM DB2 credentials file for the datamart
                        database
  --wml WML             Path to IBM WML credentials file
  --azure-studio AZURE_STUDIO
                        Path to Microsoft Azure credentials file for Microsoft
                        Azure ML Studio
  --azure-service AZURE_SERVICE
                        Path to Microsoft Azure credentials file for Microsoft
                        Azure ML Service
  --spss SPSS           Path to SPSS credentials file
  --custom CUSTOM       Path to Custom Engine credentials file
  --aws AWS             Path to Amazon Web Services credentials file
  --deployment-name DEPLOYMENT_NAME
                        Name of the existing deployment to use. Required for
                        Azure ML Studio, SPSS Engine and Custom ML Engine, but
                        optional for Watson Machine Learning. Required for
                        custom models
  --keep-schema         Use pre-existing datamart schema, only dropping all
                        tables. If not specified, datamart schema is dropped
                        and re-created
  --username USERNAME   ICP username. Required if "icp" environment is chosen,
                        not required if --iam-token is specified
  --password PASSWORD   ICP password. Required if "icp" environment is chosen,
                        not required if --iam-token is specified
  --url URL             ICP url. Required if "icp" environment is chosen
  --datamart-name DATAMART_NAME
                        Specify data mart name and database schema, default is
                        the datamart database connection username. For
                        internal database, the default is "wosfastpath"
  --history HISTORY     Days of history to preload. Default is 7
  --history-only        Store history only for existing deployment and
                        datamart. Requires --extend and --deployment-name also
                        be specified
  --history-first-day HISTORY_FIRST_DAY
                        Starting day for history. Default is 0
  --model MODEL         Sample model to set up with Watson OpenScale (default
                        "GermanCreditRiskModel")
  --list-models         Lists all available models. If a ML engine is
                        specified, then modesl specific to that engine are
                        listed
  --custom-model CUSTOM_MODEL
                        Name of custom model to set up with Watson OpenScale.
                        If specified, overrides the value set by --model. Also
                        requires that --custom-model-directory
  --custom-model-directory CUSTOM_MODEL_DIRECTORY
                        Directory with model configuration and metadata files.
                        Also requires that --custom-model be specified
  --extend              Extend existing datamart, instead of deleting and
                        recreating it
  --protect-datamart    If specified, the setup will exit if an existing
                        datamart setup is found
  --reset {metrics,monitors,datamart,model,all}
                        Reset existing datamart and/or sample models then exit
  --verbose             verbose flag
  --wml-plan {lite,standard,professional}
                        If no WML instance exists, then provision one with the
                        specified plan. Default is "lite", other plans are
                        paid plans
  --openscale-plan {lite,standard}
                        If no OpenScale instance exists, then provision one
                        with the specified plan. Default is "lite", other
                        plans are paid plans
  --version             show program's version number and exit

required arguments (only one needed):
  -a APIKEY, --apikey APIKEY
                        IBM Cloud platform user APIKey. If "--env icp" is also
                        specified, APIKey value is not used.
  -i IAM_TOKEN, --iam-token IAM_TOKEN
                        IBM Cloud authentication IAM token, or IBM Cloud
                        private authentication IAM token. Format can be
                        (--iam-token "Bearer <token>") or (--iam-token
                        <token>)

示例

在本例中，如果wml实例已经存在，则使用该实例，但如果不存在，则提供并使用新的lite计划实例。 如果存在openscale实例，则会删除其datamart并与其内部数据库架构一起重新创建。 否则，将设置lite plan openscale实例。 germancreditiskmodel存储并部署在wml中，配置为openscale，并存储7天的历史数据。 然后上传新的反馈数据，做出100个新的实时评分预测，随后进行公平性和质量检查，并给出一个解释。

exportAPIKEY=<IBM_CLOUD_API_KEY>
ibm-ai-openscale-cli --apikey $APIKEY

在本例中，假设用户已经设置了wml、openscale、ibmclouddatabaseforpostgres（icd）的实例，并为openscale datamart数据库选择了模式。 删除并重新创建openscale数据集市，删除并重新创建数据集市的数据库架构。 已部署的DrugSelectionModel实例配置为OpenScale，并存储7天的历史数据， 之后是新的反馈数据上传、100个新分数、公平性和质量检查以及一个解释。

exportAPIKEY=<IBM_CLOUD_API_KEY>
exportWML=<path to WML instance credentials JSON file>
exportICD=<path to ICD instance credentials JSON file>
exportSCHEMA=<ICD database schema name>
ibm-ai-openscale-cli --apikey $APIKEY --wml $WML --model DrugSelectionModel --deployment-name DrugSelectionModelDeployment --icd $ICD --datamart-name $SCHEMA

在本例中，假设用户已经在云上提供了ibm db2 warehouse的入口计划实例。 用户现有的db2模式中的openscale datamart表将被删除并重新创建。 germancreditiskmodel存储并部署在wml中，配置为openscale，并存储7天的历史数据， 接下来是新的反馈数据上传，100个新分数，公平性和质量检查，以及一个expl反应性：

exportAPIKEY=<IBM_CLOUD_API_KEY>
exportDB2=<path to DB2 instance credentials JSON file>
exportSCHEMA=<user's DB2 database schema>
ibm-ai-openscale-cli --apikey $APIKEY --db2 $DB2 --datamart-name $SCHEMA --keep-schema

在本例中，假设用户在wml中存储了自己的自定义模型mybusinessmodel，并将其部署为mybusinessmodeldeployment。 还假设它们已经有一个openscale的已配置实例，但该实例尚未配置。 在自定义模型目录中，用户提供了一个configuration.json文件，其中包含所需的模型配置详细信息。 创建了openscale datamart和datamart数据库模式，并将mybusinessmodeldeployment配置为openscale。 然后上载新的反馈数据（如果提供），向模型发出100个新的评分请求，然后进行公平性和质量检查（如果配置），并给出一个解释。

exportAPIKEY=<IBM_CLOUD_API_KEY>
exportWML=<path to WML instance credentials JSON file>
exportMODELPATH=<path to custom model directory>
ibm-ai-openscale-cli --apikey $APIKEY --wml $WML --custom-model MyBusinessModel --deployment-name MyBusinessModelDeployment --custom-model-directory $MODELPATH

FAQ

问：德国雷氏模型的样本模型是什么？

a.GermanCreditiskmodel示例模型取自"Watson Studio、Watson机器学习和Watson OpenScale示例" github repo，特别是ibm watson openscale教程。 当您运行ibm ai openscale cli来部署和配置germancreditiskmodel时，结果将如同您运行了适当的教程笔记本 为您的机器学习引擎。

问：凭证文件的格式是什么？

A:每个凭证文件都有自己的格式：

博士后

{
  "uri": "postgres://<USERNAME>:<PASSWORD>@<HOSTNAME>:<PORT>/<DB>"
}

IBM Postgres云数据库（ICD）

• 从IBM云中的ICD服务实例复制服务凭据

DB2

{
    "username": "<USERNAME>",
    "password": "<PASSWORD>",
    "hostname": "<HOSTNAME>",
    "port": "<PORT>",
    "db": "<DB>"
}

IBM Watson机器学习（WML）

• 从IBM云中的WML服务实例复制服务凭据

微软azure

pip install -U ibm-ai-openscale-cli

pip install -U ibm-ai-openscale-cli

pip install -U ibm-ai-openscale-cli

pip install -U ibm-ai-openscale-cli