你当前正在访问 Microsoft Azure Global Edition 技术文档网站。 如果需要访问由世纪互联运营的 Microsoft Azure 中国技术文档网站,请访问 https://docs.azure.cn

Microsoft.MachineLearningServices 工作区 2023-08-01-preview

Bicep 资源定义

可以使用针对以下操作部署工作区资源类型:

有关每个 API 版本中更改的属性的列表,请参阅 更改日志

资源格式

若要创建 Microsoft.MachineLearningServices/workspaces 资源,请将以下 Bicep 添加到模板。

resource symbolicname 'Microsoft.MachineLearningServices/workspaces@2023-08-01-preview' = {
  name: 'string'
  location: 'string'
  tags: {
    tagName1: 'tagValue1'
    tagName2: 'tagValue2'
  }
    {customized property}: 'string'
  sku: {
    capacity: int
    family: 'string'
    name: 'string'
    size: 'string'
    tier: 'string'
  }
  kind: 'string'
  identity: {
    type: 'string'
    userAssignedIdentities: {
      {customized property}: {}
    }
  }
  properties: {
    allowPublicAccessWhenBehindVnet: bool
    applicationInsights: 'string'
    associatedWorkspaces: [
      'string'
    ]
    containerRegistries: [
      'string'
    ]
    containerRegistry: 'string'
    description: 'string'
    discoveryUrl: 'string'
    enableDataIsolation: bool
    encryption: {
      cosmosDbResourceId: 'string'
      identity: {
        userAssignedIdentity: 'string'
      }
      keyVaultProperties: {
        identityClientId: 'string'
        keyIdentifier: 'string'
        keyVaultArmId: 'string'
      }
      searchAccountResourceId: 'string'
      status: 'string'
      storageAccountResourceId: 'string'
    }
    existingWorkspaces: [
      'string'
    ]
    featureStoreSettings: {
      computeRuntime: {
        sparkRuntimeVersion: 'string'
      }
      offlineStoreConnectionName: 'string'
      onlineStoreConnectionName: 'string'
    }
    friendlyName: 'string'
    hbiWorkspace: bool
    hubResourceId: 'string'
    imageBuildCompute: 'string'
    keyVault: 'string'
    keyVaults: [
      'string'
    ]
    managedNetwork: {
      isolationMode: 'string'
      outboundRules: {
        {customized property}: {
          category: 'string'
          status: 'string'
          type: 'string'
          // For remaining properties, see OutboundRule objects
        }
      }
      status: {
        sparkReady: bool
        status: 'string'
      }
    }
    primaryUserAssignedIdentity: 'string'
    publicNetworkAccess: 'string'
    serverlessComputeSettings: {
      serverlessComputeCustomSubnet: 'string'
      serverlessComputeNoPublicIP: bool
    }
    serviceManagedResourcesSettings: {
      cosmosDb: {
        collectionsThroughput: int
      }
    }
    sharedPrivateLinkResources: [
      {
        name: 'string'
        properties: {
          groupId: 'string'
          privateLinkResourceId: 'string'
          requestMessage: 'string'
          status: 'string'
        }
      }
    ]
    softDeleteRetentionInDays: int
    storageAccount: 'string'
    storageAccounts: [
      'string'
    ]
    systemDatastoresAuthMode: 'string'
    v1LegacyMode: bool
    workspaceHubConfig: {
      additionalWorkspaceStorageAccounts: [
        'string'
      ]
      defaultWorkspaceResourceGroup: 'string'
    }
  }
}

OutboundRule 对象

设置 type 属性以指定对象的类型。

对于 FQDN,请使用:

  type: 'FQDN'
  destination: 'string'

对于 PrivateEndpoint,请使用:

  type: 'PrivateEndpoint'
  destination: {
    serviceResourceId: 'string'
    sparkEnabled: bool
    sparkStatus: 'string'
    subresourceTarget: 'string'
  }

对于 ServiceTag,请使用:

  type: 'ServiceTag'
  destination: {
    action: 'string'
    portRanges: 'string'
    protocol: 'string'
    serviceTag: 'string'
  }

属性值

workspaces

名称 说明
name 资源名称 字符串 (必需)

字符限制:3-33

有效字符:
字母数字、连字符和下划线。
location 字符串
标记 {string} 的字典 标记名称和值的字典。 请参阅 模板中的标记
sku 可选。 此字段需要由 RP 实现,因为 AML 支持多个层 Sku
kind 字符串
标识 托管服务标识 (系统分配的标识和/或用户分配的标识) ManagedServiceIdentity
properties 实体的其他属性。 WorkspaceProperties (必需)

ManagedServiceIdentity

名称 说明 Value
type 托管服务标识的类型 (其中允许 systemAssigned 和 UserAssigned 类型) 。 “None”
“SystemAssigned”
“SystemAssigned,UserAssigned”
“UserAssigned” (必需)
userAssignedIdentities 与资源关联的用户分配标识集。 userAssignedIdentities 字典密钥将是采用以下格式的 ARM 资源 ID:'/subscriptions/{subscriptionId}/resourceGroups/{resourceGroupName}/providers/Microsoft.ManagedIdentity/userAssignedIdentities/{identityName}。 字典值可以是请求中 ({}) 的空对象。 UserAssignedIdentities

UserAssignedIdentities

名称 说明
{customized property} UserAssignedIdentity

UserAssignedIdentity

此对象不包含在部署期间要设置的任何属性。 所有属性均为 ReadOnly。

WorkspaceProperties

名称 说明
allowPublicAccessWhenBehindVnet 指示在 VNet 后面时是否允许公共访问的标志。 bool
applicationInsights 与此工作区关联的应用程序见解的 ARM ID。 string
associatedWorkspaces string[]
containerRegistries string[]
containerRegistry 与此工作区关联的容器注册表的 ARM ID。 字符串
description 此工作区的说明。 string
discoveryUrl 发现服务的 URL,用于标识机器学习试验服务的区域终结点 string
enableDataIsolation bool
加密 EncryptionProperty
existingWorkspaces string[]
featureStoreSettings 功能存储类型工作区的设置。 FeatureStoreSettings
friendlyName 此工作区的友好名称。 此名称以可变为单位 字符串
hbiWorkspace 用于向工作区中的 HBI 数据发出信号并减少服务收集的诊断数据的标志 bool
hubResourceId 字符串
imageBuildCompute 映像生成的计算名称 字符串
keyVault 与此工作区关联的密钥保管库的 ARM ID。 创建工作区后,无法更改此情况 字符串
keyVaults string[]
managedNetwork 机器学习工作区的托管网络设置。 ManagedNetworkSettings
primaryUserAssignedIdentity 表示工作区标识的用户分配的标识资源 ID。 字符串
publicNetworkAccess 是否允许来自公用网络的请求。 “Disabled”
“已启用”
serverlessComputeSettings 在工作区中创建的无服务器计算的设置 ServerlessComputeSettings
serviceManagedResourcesSettings 服务托管的资源设置。 ServiceManagedResourcesSettings
sharedPrivateLinkResources 此工作区中的共享专用链接资源列表。 SharedPrivateLinkResource[]
softDeleteRetentionInDays 软删除工作区后的保留时间(以天为单位)。 int
storageAccount 与此工作区关联的存储帐户的 ARM ID。 创建工作区后,无法更改此情况 字符串
storageAccounts string[]
systemDatastoresAuthMode 用于访问工作区的系统数据存储的身份验证模式。 字符串
v1LegacyMode 启用 v1_legacy_mode 可能会阻止你使用 v2 API 提供的功能。 bool
workspaceHubConfig WorkspaceHub 的配置对象。 WorkspaceHubConfig

EncryptionProperty

名称 说明
cosmosDbResourceId 客户用于存储客户数据的 byok cosmosdb 帐户
使用加密
字符串
标识 要与 keyVault 一起使用的标识 IdentityForCmk
keyVaultProperties 用于加密的 KeyVault 详细信息 需要 KeyVaultProperties ()
searchAccountResourceId 客户用于存储客户数据的 byok 搜索帐户
使用加密
字符串
status 指示是否为工作区启用加密。 “Disabled”
“已启用” (必需)
storageAccountResourceId 客户用于存储客户数据的 byok 存储帐户
使用加密
字符串

IdentityForCmk

名称 说明
userAssignedIdentity 用于从 keyVault 提取加密密钥的 UserAssignedIdentity 字符串

KeyVaultProperties

名称 说明
identityClientId 目前,我们仅支持 SystemAssigned MSI。
当我们支持 UserAssignedIdentities 时,我们需要此
字符串
keyIdentifier 用于加密数据的 KeyVault 密钥标识符 字符串 (必需)

约束:
最小长度 = 1
Pattern = [a-zA-Z0-9_]
keyVaultArmId 包含数据加密密钥的 KeyVault Arm ID 字符串 (必需)

约束:
最小长度 = 1
Pattern = [a-zA-Z0-9_]

FeatureStoreSettings

名称 说明
computeRuntime ComputeRuntimeDto
offlineStoreConnectionName 字符串
onlineStoreConnectionName 字符串

ComputeRuntimeDto

名称 说明
sparkRuntimeVersion string

ManagedNetworkSettings

名称 说明
isolationMode 机器学习工作区托管网络的隔离模式。 “AllowInternetOutbound”
“AllowOnlyApprovedOutbound”
“Disabled”
outboundRules {OutboundRule} 的字典 ManagedNetworkSettingsOutboundRules
status 机器学习工作区托管网络的预配状态。 ManagedNetworkProvisionStatus

ManagedNetworkSettingsOutboundRules

名称 说明
{customized property} OutboundRule

OutboundRule

名称 说明
category 机器学习工作区的托管网络出站规则的类别。 “建议”
“必需”
“UserDefined”
status 机器学习工作区的托管网络出站规则的类型。 “Active”
“非活动”
类型 设置对象类型 FQDN
PrivateEndpoint
需要 serviceTag ()

FqdnOutboundRule

名称 说明 Value
type 机器学习工作区的托管网络出站规则的类型。 “FQDN” (必需)
destination 字符串

PrivateEndpointOutboundRule

名称 说明 Value
type 机器学习工作区的托管网络出站规则的类型。 “PrivateEndpoint” (必需)
destination 机器学习工作区托管网络的专用终结点出站规则的专用终结点目标。 PrivateEndpointDestination

PrivateEndpointDestination

名称 说明
serviceResourceId 字符串
sparkEnabled bool
sparkStatus 机器学习工作区的托管网络出站规则的类型。 “Active”
“非活动”
subresourceTarget 字符串

ServiceTagOutboundRule

名称 说明 Value
type 机器学习工作区的托管网络出站规则的类型。 “ServiceTag” (必需)
destination 机器学习工作区托管网络的服务标记出站规则的服务标记目标。 ServiceTagDestination

ServiceTagDestination

名称 说明
action 网络规则的操作枚举。 “允许”
“拒绝”
portRanges 字符串
protocol 字符串
serviceTag 字符串

ManagedNetworkProvisionStatus

名称 说明
sparkReady bool
status 机器学习工作区的托管网络的状态。 “Active”
“非活动”

ServerlessComputeSettings

名称 说明
serverlessComputeCustomSubnet 应在其中部署无服务器计算节点的现有虚拟网络子网的资源 ID 字符串
serverlessComputeNoPublicIP 指示是否在自定义 vNet 中部署的无服务器计算节点对于具有专用终结点的工作区没有公共 IP 地址的标志 bool

ServiceManagedResourcesSettings

名称 说明
CosmosDB CosmosDbSettings

CosmosDbSettings

名称 说明
collectionsThroughput int

SharedPrivateLinkResource

名称 说明
name 专用链接的唯一名称 string
properties 共享专用链接资源的属性。 SharedPrivateLinkResourceProperty

SharedPrivateLinkResourceProperty

名称 说明
groupId 专用链接的组 ID string
privateLinkResourceId 专用链接链接到的资源 ID string
requestMessage 请求消息 字符串
status 服务使用者与服务提供商的连接状态 “已批准”
“Disconnected”
'Pending'
“Rejected”
“Timeout”

WorkspaceHubConfig

名称 说明
additionalWorkspaceStorageAccounts string[]
defaultWorkspaceResourceGroup string

SKU

名称 说明
容量 如果 SKU 支持横向扩展/缩减,则应包含容量整数。 如果无法对资源进行横向扩展/缩减,则可能会省略此项。 int
family 如果服务针对同一 SKU 具有不同代系的硬件,则可以在此处捕获。 string
name SKU 的名称。 例如 - P3。 它通常是字母+数字代码 字符串 (必需)
大小 SKU 大小。 当名称字段是层和一些其他值的组合时,这将是独立代码。 字符串
如果服务具有多个层,但 PUT 上不需要此字段,则资源提供程序需要实现此字段。 'Basic'
“免费”
“Premium”
'Standard'

快速入门模板

以下快速入门模板部署此资源类型。

模板 说明
Azure 机器学习工作区

部署到 Azure
此模板创建新的 Azure 机器学习工作区,以及加密的存储帐户、KeyVault 和 Applications Insights 日志记录
创建具有多个数据集的 AML 工作区 & 数据存储

部署到 Azure
此模板创建包含多个数据集的 Azure 机器学习工作区 & 数据存储。
Azure 机器学习端到端安全设置

部署到 Azure
这组 Bicep 模板演示如何在安全设置中端到端地设置 Azure 机器学习。 此参考实现包括工作区、计算群集、计算实例和附加的专用 AKS 群集。
Azure 机器学习端到端安全设置 (旧版)

部署到 Azure
这组 Bicep 模板演示如何在安全设置中端到端地设置 Azure 机器学习。 此参考实现包括工作区、计算群集、计算实例和附加的专用 AKS 群集。
创建具有专用 IP 地址的 AKS 计算目标

部署到 Azure
此模板在给定的 Azure 机器学习服务工作区中创建具有专用 IP 地址的 AKS 计算目标。
创建 Azure 机器学习服务工作区

部署到 Azure
此部署模板指定 Azure 机器学习工作区及其关联的资源,包括 Azure 密钥保管库、Azure 存储、Azure 应用程序见解和Azure 容器注册表。 此配置描述了开始使用 Azure 机器学习所需的最少资源集。
(CMK) 创建 Azure 机器学习服务工作区

部署到 Azure
此部署模板指定 Azure 机器学习工作区及其关联的资源,包括 Azure 密钥保管库、Azure 存储、Azure 应用程序见解和Azure 容器注册表。 该示例演示如何使用客户管理的加密密钥配置 Azure 机器学习进行加密。
(vnet) 创建 Azure 机器学习服务工作区

部署到 Azure
此部署模板指定 Azure 机器学习工作区及其关联的资源,包括 Azure 密钥保管库、Azure 存储、Azure 应用程序见解和Azure 容器注册表。 此配置描述了在网络隔离设置中开始使用 Azure 机器学习所需的资源集。
(旧版) 创建 Azure 机器学习服务工作区

部署到 Azure
此部署模板指定 Azure 机器学习工作区及其关联的资源,包括 Azure 密钥保管库、Azure 存储、Azure 应用程序见解和Azure 容器注册表。 此配置描述了在网络隔离设置中开始使用 Azure 机器学习所需的资源集。

ARM 模板资源定义

可以使用面向以下操作部署工作区资源类型:

有关每个 API 版本中已更改属性的列表,请参阅 更改日志

资源格式

若要创建 Microsoft.MachineLearningServices/workspaces 资源,请将以下 JSON 添加到模板。

{
  "type": "Microsoft.MachineLearningServices/workspaces",
  "apiVersion": "2023-08-01-preview",
  "name": "string",
  "location": "string",
  "tags": {
    "tagName1": "tagValue1",
    "tagName2": "tagValue2"
  },
    "{customized property}": "string"
  "sku": {
    "capacity": "int",
    "family": "string",
    "name": "string",
    "size": "string",
    "tier": "string"
  },
  "kind": "string",
  "identity": {
    "type": "string",
    "userAssignedIdentities": {
      "{customized property}": {}
    }
  },
  "properties": {
    "allowPublicAccessWhenBehindVnet": "bool",
    "applicationInsights": "string",
    "associatedWorkspaces": [ "string" ],
    "containerRegistries": [ "string" ],
    "containerRegistry": "string",
    "description": "string",
    "discoveryUrl": "string",
    "enableDataIsolation": "bool",
    "encryption": {
      "cosmosDbResourceId": "string",
      "identity": {
        "userAssignedIdentity": "string"
      },
      "keyVaultProperties": {
        "identityClientId": "string",
        "keyIdentifier": "string",
        "keyVaultArmId": "string"
      },
      "searchAccountResourceId": "string",
      "status": "string",
      "storageAccountResourceId": "string"
    },
    "existingWorkspaces": [ "string" ],
    "featureStoreSettings": {
      "computeRuntime": {
        "sparkRuntimeVersion": "string"
      },
      "offlineStoreConnectionName": "string",
      "onlineStoreConnectionName": "string"
    },
    "friendlyName": "string",
    "hbiWorkspace": "bool",
    "hubResourceId": "string",
    "imageBuildCompute": "string",
    "keyVault": "string",
    "keyVaults": [ "string" ],
    "managedNetwork": {
      "isolationMode": "string",
      "outboundRules": {
        "{customized property}": {
          "category": "string",
          "status": "string",
          "type": "string"
          // For remaining properties, see OutboundRule objects
        }
      },
      "status": {
        "sparkReady": "bool",
        "status": "string"
      }
    },
    "primaryUserAssignedIdentity": "string",
    "publicNetworkAccess": "string",
    "serverlessComputeSettings": {
      "serverlessComputeCustomSubnet": "string",
      "serverlessComputeNoPublicIP": "bool"
    },
    "serviceManagedResourcesSettings": {
      "cosmosDb": {
        "collectionsThroughput": "int"
      }
    },
    "sharedPrivateLinkResources": [
      {
        "name": "string",
        "properties": {
          "groupId": "string",
          "privateLinkResourceId": "string",
          "requestMessage": "string",
          "status": "string"
        }
      }
    ],
    "softDeleteRetentionInDays": "int",
    "storageAccount": "string",
    "storageAccounts": [ "string" ],
    "systemDatastoresAuthMode": "string",
    "v1LegacyMode": "bool",
    "workspaceHubConfig": {
      "additionalWorkspaceStorageAccounts": [ "string" ],
      "defaultWorkspaceResourceGroup": "string"
    }
  }
}

OutboundRule 对象

设置 type 属性以指定对象的类型。

对于 FQDN,请使用:

  "type": "FQDN",
  "destination": "string"

对于 PrivateEndpoint,请使用:

  "type": "PrivateEndpoint",
  "destination": {
    "serviceResourceId": "string",
    "sparkEnabled": "bool",
    "sparkStatus": "string",
    "subresourceTarget": "string"
  }

对于 ServiceTag,请使用:

  "type": "ServiceTag",
  "destination": {
    "action": "string",
    "portRanges": "string",
    "protocol": "string",
    "serviceTag": "string"
  }

属性值

workspaces

名称 说明 Value
type 资源类型 “Microsoft.MachineLearningServices/workspaces”
apiVersion 资源 API 版本 “2023-08-01-preview”
name 资源名称 字符串 (必需)

字符限制:3-33

有效字符:
字母数字、连字符和下划线字符。
location 字符串
标记 {string} 的字典 标记名称和值的字典。 请参阅 模板中的标记
sku 可选。 此字段需要由 RP 实现,因为 AML 支持多个层 Sku
kind 字符串
标识 托管服务标识 (系统分配的标识和/或用户分配的标识) ManagedServiceIdentity
properties 实体的其他属性。 WorkspaceProperties (必需)

ManagedServiceIdentity

名称 说明 Value
type 托管服务标识的类型 (其中允许 systemAssigned 和 UserAssigned 类型) 。 “None”
“SystemAssigned”
“SystemAssigned,UserAssigned”
“UserAssigned” (必需)
userAssignedIdentities 与资源关联的用户分配标识集。 userAssignedIdentities 字典密钥将是采用以下格式的 ARM 资源 ID:'/subscriptions/{subscriptionId}/resourceGroups/{resourceGroupName}/providers/Microsoft.ManagedIdentity/userAssignedIdentities/{identityName}。 字典值可以是请求中 ({}) 的空对象。 UserAssignedIdentities

UserAssignedIdentities

名称 说明
{customized property} UserAssignedIdentity

UserAssignedIdentity

此对象不包含在部署期间要设置的任何属性。 所有属性均为 ReadOnly。

WorkspaceProperties

名称 说明
allowPublicAccessWhenBehindVnet 指示在 VNet 后面时是否允许公共访问的标志。 bool
applicationInsights 与此工作区关联的应用程序见解的 ARM ID。 字符串
associatedWorkspaces string[]
containerRegistries string[]
containerRegistry 与此工作区关联的容器注册表的 ARM ID。 string
description 此工作区的说明。 字符串
discoveryUrl 发现服务的 URL,用于标识机器学习试验服务的区域终结点 字符串
enableDataIsolation bool
加密 EncryptionProperty
existingWorkspaces string[]
featureStoreSettings 功能存储类型工作区的设置。 FeatureStoreSettings
friendlyName 此工作区的友好名称。 此名称以可变为单位 字符串
hbiWorkspace 用于向工作区中的 HBI 数据发出信号并减少服务收集的诊断数据的标志 bool
hubResourceId string
imageBuildCompute 映像生成的计算名称 字符串
keyVault 与此工作区关联的密钥保管库的 ARM ID。 创建工作区后,无法更改此情况 字符串
keyVaults string[]
managedNetwork 机器学习工作区的托管网络设置。 ManagedNetworkSettings
primaryUserAssignedIdentity 表示工作区标识的用户分配的标识资源 ID。 string
publicNetworkAccess 是否允许来自公用网络的请求。 'Disabled'
'Enabled'
serverlessComputeSettings 在工作区中创建的无服务器计算设置 ServerlessComputeSettings
serviceManagedResourcesSettings 服务托管的资源设置。 ServiceManagedResourcesSettings
sharedPrivateLinkResources 此工作区中的共享专用链接资源列表。 SharedPrivateLinkResource[]
softDeleteRetentionInDays 工作区软删除后的保留时间(以天为单位)。 int
storageAccount 与此工作区关联的存储帐户的 ARM ID。 创建工作区后,无法更改此情况 string
storageAccounts string[]
systemDatastoresAuthMode 用于访问工作区的系统数据存储的身份验证模式。 字符串
v1LegacyMode 启用 v1_legacy_mode 可能会阻止你使用 v2 API 提供的功能。 bool
workspaceHubConfig WorkspaceHub 的配置对象。 WorkspaceHubConfig

EncryptionProperty

名称 说明
cosmosDbResourceId 客户用于存储客户数据的 byok cosmosdb 帐户
使用加密
字符串
标识 要与 keyVault 一起使用的标识 IdentityForCmk
keyVaultProperties 用于加密的 KeyVault 详细信息 需要 keyVaultProperties ()
searchAccountResourceId 客户用于存储客户数据的 byok 搜索帐户
使用加密
字符串
status 指示是否为工作区启用加密。 'Disabled'
“Enabled” (必需)
storageAccountResourceId 客户用于存储客户数据的 byok 存储帐户
使用加密
字符串

IdentityForCmk

名称 说明
userAssignedIdentity 用于从 keyVault 提取加密密钥的 UserAssignedIdentity 字符串

KeyVaultProperties

名称 说明
identityClientId 目前,我们仅支持 SystemAssigned MSI。
当我们支持 UserAssignedIdentities 时,我们需要此
string
keyIdentifier 用于加密数据的 KeyVault 密钥标识符 字符串 (必需)

约束:
最小长度 = 1
模式 = [a-zA-Z0-9_]
keyVaultArmId 包含数据加密密钥的 KeyVault Arm ID 字符串 (必需)

约束:
最小长度 = 1
模式 = [a-zA-Z0-9_]

FeatureStoreSettings

名称 说明
computeRuntime ComputeRuntimeDto
offlineStoreConnectionName 字符串
onlineStoreConnectionName string

ComputeRuntimeDto

名称 说明
sparkRuntimeVersion 字符串

ManagedNetworkSettings

名称 说明
isolationMode 机器学习工作区托管网络的隔离模式。 “AllowInternetOutbound”
“AllowOnlyApprovedOutbound”
“Disabled”
outboundRules {OutboundRule} 的字典 ManagedNetworkSettingsOutboundRules
status 机器学习工作区托管网络的预配状态。 ManagedNetworkProvisionStatus

ManagedNetworkSettingsOutboundRules

名称 说明
{customized property} OutboundRule

OutboundRule

名称 说明
category 机器学习工作区的托管网络出站规则的类别。 “建议”
“必需”
“UserDefined”
status 机器学习工作区的托管网络出站规则的类型。 “Active”
“非活动”
类型 设置对象类型 FQDN
PrivateEndpoint
需要 serviceTag ()

FqdnOutboundRule

名称 说明 Value
type 机器学习工作区的托管网络出站规则的类型。 “FQDN” (必需)
destination 字符串

PrivateEndpointOutboundRule

名称 说明 Value
type 机器学习工作区的托管网络出站规则的类型。 “PrivateEndpoint” (必需)
destination 机器学习工作区托管网络的专用终结点出站规则的专用终结点目标。 PrivateEndpointDestination

PrivateEndpointDestination

名称 说明
serviceResourceId 字符串
sparkEnabled bool
sparkStatus 机器学习工作区的托管网络出站规则的类型。 “Active”
“非活动”
subresourceTarget 字符串

ServiceTagOutboundRule

名称 说明 Value
type 机器学习工作区的托管网络出站规则的类型。 “ServiceTag” (必需)
destination 机器学习工作区托管网络的服务标记出站规则的服务标记目标。 ServiceTagDestination

ServiceTagDestination

名称 说明
action 网络规则的操作枚举。 “允许”
“拒绝”
portRanges 字符串
protocol 字符串
serviceTag 字符串

ManagedNetworkProvisionStatus

名称 说明
sparkReady bool
status 机器学习工作区的托管网络的状态。 “Active”
“非活动”

ServerlessComputeSettings

名称 说明
serverlessComputeCustomSubnet 应在其中部署无服务器计算节点的现有虚拟网络子网的资源 ID string
serverlessComputeNoPublicIP 指示是否在自定义 vNet 中部署的无服务器计算节点对于具有专用终结点的工作区没有公共 IP 地址的标志 bool

ServiceManagedResourcesSettings

名称 说明
CosmosDB CosmosDbSettings

CosmosDbSettings

名称 说明
collectionsThroughput int

SharedPrivateLinkResource

名称 说明
name 专用链接的唯一名称 string
properties 共享专用链接资源的属性。 SharedPrivateLinkResourceProperty

SharedPrivateLinkResourceProperty

名称 说明
groupId 专用链接的组 ID 字符串
privateLinkResourceId 专用链接链接到的资源 ID 字符串
requestMessage 请求消息 字符串
status 服务使用者与服务提供商的连接状态 “已批准”
“Disconnected”
“挂起”
“Rejected”
“Timeout”

WorkspaceHubConfig

名称 说明
additionalWorkspaceStorageAccounts string[]
defaultWorkspaceResourceGroup string

SKU

名称 说明
容量 如果 SKU 支持横向扩展/缩减,则应包含容量整数。 如果资源无法横向扩展/缩减,则可能会省略这一点。 int
family 如果服务具有不同代系的硬件(对于同一 SKU),则可以在此处进行捕获。 string
name SKU 的名称。 例如 - P3。 它通常是字母+数字代码 字符串 (必需)
大小 SKU 大小。 当名称字段是层和其他某个值的组合时,这将是独立代码。 字符串
如果服务具有多个层,但 PUT 上不需要此字段,则资源提供程序必须实现此字段。 “Basic”
“免费”
“Premium”
'Standard'

快速入门模板

以下快速入门模板部署此资源类型。

模板 说明
Azure 机器学习工作区

部署到 Azure
此模板创建新的 Azure 机器学习工作区,以及加密的存储帐户、KeyVault 和 Applications Insights 日志记录
创建具有多个数据集的 AML 工作区 & 数据存储

部署到 Azure
此模板创建包含多个数据集的 Azure 机器学习工作区 & 数据存储。
Azure 机器学习端到端安全设置

部署到 Azure
这组 Bicep 模板演示如何在安全设置中端到端地设置 Azure 机器学习。 此参考实现包括工作区、计算群集、计算实例和附加的专用 AKS 群集。
Azure 机器学习端到端安全设置 (旧版)

部署到 Azure
这组 Bicep 模板演示如何在安全设置中端到端地设置 Azure 机器学习。 此参考实现包括工作区、计算群集、计算实例和附加的专用 AKS 群集。
创建具有专用 IP 地址的 AKS 计算目标

部署到 Azure
此模板在给定的 Azure 机器学习服务工作区中创建具有专用 IP 地址的 AKS 计算目标。
创建 Azure 机器学习服务工作区

部署到 Azure
此部署模板指定 Azure 机器学习工作区及其关联的资源,包括 Azure 密钥保管库、Azure 存储、Azure 应用程序见解和Azure 容器注册表。 此配置描述了开始使用 Azure 机器学习所需的最少资源集。
(CMK) 创建 Azure 机器学习服务工作区

部署到 Azure
此部署模板指定 Azure 机器学习工作区及其关联的资源,包括 Azure 密钥保管库、Azure 存储、Azure 应用程序见解和Azure 容器注册表。 该示例演示如何使用客户管理的加密密钥配置 Azure 机器学习进行加密。
(vnet) 创建 Azure 机器学习服务工作区

部署到 Azure
此部署模板指定 Azure 机器学习工作区及其关联的资源,包括 Azure 密钥保管库、Azure 存储、Azure 应用程序见解和Azure 容器注册表。 此配置描述了在网络隔离设置中开始使用 Azure 机器学习所需的资源集。
(旧版) 创建 Azure 机器学习服务工作区

部署到 Azure
此部署模板指定 Azure 机器学习工作区及其关联的资源,包括 Azure 密钥保管库、Azure 存储、Azure 应用程序见解和Azure 容器注册表。 此配置描述了在网络隔离设置中开始使用 Azure 机器学习所需的资源集。

Terraform (AzAPI 提供程序) 资源定义

可以使用面向以下操作部署工作区资源类型:

  • 资源组

有关每个 API 版本中已更改属性的列表,请参阅 更改日志

资源格式

若要创建 Microsoft.MachineLearningServices/workspaces 资源,请将以下 Terraform 添加到模板。

resource "azapi_resource" "symbolicname" {
  type = "Microsoft.MachineLearningServices/workspaces@2023-08-01-preview"
  name = "string"
  location = "string"
  parent_id = "string"
  tags = {
    tagName1 = "tagValue1"
    tagName2 = "tagValue2"
  }
    {customized property} = "string"
  identity {
    type = "string"
    identity_ids = []
  }
  body = jsonencode({
    properties = {
      allowPublicAccessWhenBehindVnet = bool
      applicationInsights = "string"
      associatedWorkspaces = [
        "string"
      ]
      containerRegistries = [
        "string"
      ]
      containerRegistry = "string"
      description = "string"
      discoveryUrl = "string"
      enableDataIsolation = bool
      encryption = {
        cosmosDbResourceId = "string"
        identity = {
          userAssignedIdentity = "string"
        }
        keyVaultProperties = {
          identityClientId = "string"
          keyIdentifier = "string"
          keyVaultArmId = "string"
        }
        searchAccountResourceId = "string"
        status = "string"
        storageAccountResourceId = "string"
      }
      existingWorkspaces = [
        "string"
      ]
      featureStoreSettings = {
        computeRuntime = {
          sparkRuntimeVersion = "string"
        }
        offlineStoreConnectionName = "string"
        onlineStoreConnectionName = "string"
      }
      friendlyName = "string"
      hbiWorkspace = bool
      hubResourceId = "string"
      imageBuildCompute = "string"
      keyVault = "string"
      keyVaults = [
        "string"
      ]
      managedNetwork = {
        isolationMode = "string"
        outboundRules = {
          {customized property} = {
            category = "string"
            status = "string"
            type = "string"
            // For remaining properties, see OutboundRule objects
          }
        }
        status = {
          sparkReady = bool
          status = "string"
        }
      }
      primaryUserAssignedIdentity = "string"
      publicNetworkAccess = "string"
      serverlessComputeSettings = {
        serverlessComputeCustomSubnet = "string"
        serverlessComputeNoPublicIP = bool
      }
      serviceManagedResourcesSettings = {
        cosmosDb = {
          collectionsThroughput = int
        }
      }
      sharedPrivateLinkResources = [
        {
          name = "string"
          properties = {
            groupId = "string"
            privateLinkResourceId = "string"
            requestMessage = "string"
            status = "string"
          }
        }
      ]
      softDeleteRetentionInDays = int
      storageAccount = "string"
      storageAccounts = [
        "string"
      ]
      systemDatastoresAuthMode = "string"
      v1LegacyMode = bool
      workspaceHubConfig = {
        additionalWorkspaceStorageAccounts = [
          "string"
        ]
        defaultWorkspaceResourceGroup = "string"
      }
    }
    sku = {
      capacity = int
      family = "string"
      name = "string"
      size = "string"
      tier = "string"
    }
    kind = "string"
  })
}

OutboundRule 对象

设置 type 属性以指定对象的类型。

对于 FQDN,请使用:

  type = "FQDN"
  destination = "string"

对于 PrivateEndpoint,请使用:

  type = "PrivateEndpoint"
  destination = {
    serviceResourceId = "string"
    sparkEnabled = bool
    sparkStatus = "string"
    subresourceTarget = "string"
  }

对于 ServiceTag,请使用:

  type = "ServiceTag"
  destination = {
    action = "string"
    portRanges = "string"
    protocol = "string"
    serviceTag = "string"
  }

属性值

workspaces

名称 说明 Value
type 资源类型 “Microsoft.MachineLearningServices/workspaces@2023-08-01-preview”
name 资源名称 字符串 (必需)

字符限制:3-33

有效字符:
字母数字、连字符和下划线字符。
location 字符串
parent_id 若要部署到资源组,请使用该资源组的 ID。 字符串 (必需)
标记 {string} 字典 标记名称和值的字典。
sku 可选。 此字段需要由 RP 实现,因为 AML 支持多个层 Sku
kind 字符串
标识 托管服务标识 (系统分配的标识和/或用户分配的标识) ManagedServiceIdentity
properties 实体的其他属性。 WorkspaceProperties (必需)

ManagedServiceIdentity

名称 说明 Value
type ) 允许 SystemAssigned 和 UserAssigned 类型的托管服务标识 (的类型。 “SystemAssigned”
“SystemAssigned,UserAssigned”
“UserAssigned” (必需)
identity_ids 与资源关联的用户分配标识集。 userAssignedIdentities 字典密钥将是采用以下格式的 ARM 资源 ID:'/subscriptions/{subscriptionId}/resourceGroups/{resourceGroupName}/providers/Microsoft.ManagedIdentity/userAssignedIdentities/{identityName}。 字典值可以是请求中) ({} 空对象。 用户标识 ID 的数组。

UserAssignedIdentities

名称 说明
{customized property} UserAssignedIdentity

UserAssignedIdentity

此对象不包含在部署期间要设置的任何属性。 所有属性均为 ReadOnly。

WorkspaceProperties

名称 说明
allowPublicAccessWhenBehindVnet 指示在 VNet 后面时是否允许公共访问的标志。 bool
applicationInsights 与此工作区关联的应用程序见解的 ARM ID。 字符串
associatedWorkspaces string[]
containerRegistries string[]
containerRegistry 与此工作区关联的容器注册表的 ARM ID。 string
description 此工作区的说明。 字符串
discoveryUrl 发现服务的 URL,用于标识机器学习试验服务的区域终结点 字符串
enableDataIsolation bool
加密 EncryptionProperty
existingWorkspaces string[]
featureStoreSettings 功能存储类型工作区的设置。 FeatureStoreSettings
friendlyName 此工作区的友好名称。 此名称以可变为单位 字符串
hbiWorkspace 用于向工作区中的 HBI 数据发出信号并减少服务收集的诊断数据的标志 bool
hubResourceId string
imageBuildCompute 映像生成的计算名称 string
keyVault 与此工作区关联的密钥保管库的 ARM ID。 创建工作区后,无法更改此情况 字符串
keyVaults string[]
managedNetwork 机器学习工作区的托管网络设置。 ManagedNetworkSettings
primaryUserAssignedIdentity 表示工作区标识的用户分配的标识资源 ID。 字符串
publicNetworkAccess 是否允许来自公用网络的请求。 “Disabled”
“已启用”
serverlessComputeSettings 在工作区中创建的无服务器计算的设置 ServerlessComputeSettings
serviceManagedResourcesSettings 服务托管的资源设置。 ServiceManagedResourcesSettings
sharedPrivateLinkResources 此工作区中的共享专用链接资源列表。 SharedPrivateLinkResource[]
softDeleteRetentionInDays 软删除工作区后的保留时间(以天为单位)。 int
storageAccount 与此工作区关联的存储帐户的 ARM ID。 创建工作区后,无法更改此情况 string
storageAccounts string[]
systemDatastoresAuthMode 用于访问工作区的系统数据存储的身份验证模式。 string
v1LegacyMode 启用 v1_legacy_mode 可能会阻止你使用 v2 API 提供的功能。 bool
workspaceHubConfig WorkspaceHub 的配置对象。 WorkspaceHubConfig

EncryptionProperty

名称 说明
cosmosDbResourceId 客户用于存储客户数据的 byok cosmosdb 帐户
使用加密
字符串
标识 要与 keyVault 一起使用的标识 IdentityForCmk
keyVaultProperties 用于加密的 KeyVault 详细信息 需要 KeyVaultProperties ()
searchAccountResourceId 客户用于存储客户数据的 byok 搜索帐户
使用加密
字符串
status 指示是否为工作区启用加密。 “Disabled”
“已启用” (必需)
storageAccountResourceId 客户用于存储客户数据的 byok 存储帐户
使用加密
string

IdentityForCmk

名称 说明
userAssignedIdentity 用于从 keyVault 提取加密密钥的 UserAssignedIdentity 字符串

KeyVaultProperties

名称 说明
identityClientId 目前,我们仅支持 SystemAssigned MSI。
当我们支持 UserAssignedIdentities 时,我们需要此
字符串
keyIdentifier 用于加密数据的 KeyVault 密钥标识符 字符串 (必需)

约束:
最小长度 = 1
Pattern = [a-zA-Z0-9_]
keyVaultArmId 包含数据加密密钥的 KeyVault Arm ID 字符串 (必需)

约束:
最小长度 = 1
Pattern = [a-zA-Z0-9_]

FeatureStoreSettings

名称 说明
computeRuntime ComputeRuntimeDto
offlineStoreConnectionName 字符串
onlineStoreConnectionName 字符串

ComputeRuntimeDto

名称 说明
sparkRuntimeVersion 字符串

ManagedNetworkSettings

名称 说明
isolationMode 机器学习工作区托管网络的隔离模式。 “AllowInternetOutbound”
“AllowOnlyApprovedOutbound”
“Disabled”
outboundRules {OutboundRule} 的字典 ManagedNetworkSettingsOutboundRules
status 机器学习工作区托管网络的预配状态。 ManagedNetworkProvisionStatus

ManagedNetworkSettingsOutboundRules

名称 说明
{customized property} OutboundRule

OutboundRule

名称 说明
category 机器学习工作区的托管网络出站规则的类别。 “建议”
“必需”
“UserDefined”
status 机器学习工作区的托管网络出站规则的类型。 “活动”
“非活动”
类型 设置对象类型 FQDN
PrivateEndpoint
需要 serviceTag ()

FqdnOutboundRule

名称 说明 Value
type 机器学习工作区的托管网络出站规则的类型。 “FQDN” (必需)
destination 字符串

PrivateEndpointOutboundRule

名称 说明 Value
type 机器学习工作区的托管网络出站规则的类型。 “PrivateEndpoint” (必需)
destination 机器学习工作区托管网络的专用终结点出站规则的专用终结点目标。 PrivateEndpointDestination

PrivateEndpointDestination

名称 说明
serviceResourceId 字符串
sparkEnabled bool
sparkStatus 机器学习工作区的托管网络出站规则的类型。 “活动”
“非活动”
subresourceTarget 字符串

ServiceTagOutboundRule

名称 说明 Value
type 机器学习工作区的托管网络出站规则的类型。 “ServiceTag” (必需)
destination 机器学习工作区托管网络的服务标记出站规则的服务标记目标。 ServiceTagDestination

ServiceTagDestination

名称 说明
action 网络规则的操作枚举。 “允许”
“拒绝”
portRanges string
protocol 字符串
serviceTag 字符串

ManagedNetworkProvisionStatus

名称 说明
sparkReady bool
status 机器学习工作区的托管网络的状态。 “活动”
“非活动”

ServerlessComputeSettings

名称 说明
serverlessComputeCustomSubnet 应在其中部署无服务器计算节点的现有虚拟网络子网的资源 ID string
serverlessComputeNoPublicIP 指示是否在自定义 vNet 中部署的无服务器计算节点对于具有专用终结点的工作区没有公共 IP 地址的标志 bool

ServiceManagedResourcesSettings

名称 说明
CosmosDB CosmosDbSettings

CosmosDbSettings

名称 说明
collectionsThroughput int

SharedPrivateLinkResource

名称 说明
name 专用链接的唯一名称 string
properties 共享专用链接资源的属性。 SharedPrivateLinkResourceProperty

SharedPrivateLinkResourceProperty

名称 说明
groupId 专用链接的组 ID 字符串
privateLinkResourceId 专用链接链接到的资源 ID 字符串
requestMessage 请求消息 字符串
status 服务使用者与服务提供商的连接状态 “已批准”
“已断开连接”
“挂起”
“已拒绝”
“Timeout”

WorkspaceHubConfig

名称 说明
additionalWorkspaceStorageAccounts string[]
defaultWorkspaceResourceGroup 字符串

SKU

名称 说明
容量 如果 SKU 支持横向扩展/缩减,则应包含容量整数。 如果资源无法横向扩展/缩减,则可能会省略这一点。 int
family 如果服务具有不同代系的硬件(对于同一 SKU),则可以在此处进行捕获。 string
name SKU 的名称。 例如 - P3。 它通常是字母+数字代码 字符串 (必需)
大小 SKU 大小。 当名称字段是层和其他某个值的组合时,这将是独立代码。 字符串
如果服务具有多个层,但 PUT 上不需要此字段,则资源提供程序必须实现此字段。 "Basic"
"Free"
“Premium”
“Standard”