ZStack Logo

ZStack AIOS

部署模型评测任务 (DeployModelEvalService)

面向开发者的 API 使用规范、SDK 调用方式和 AIOS 相关接口说明。

API请求

URLs
PUT zstack/v1/ai/model-services/eval/{uuid}
Headers
Authorization: OAuth the-session-uuid
Body
{
  "deployModelEvalService": {
    "taskType": "Performance",
    "limits": 10,
    "temperature": 0.5,
    "topK": 10,
    "topP": 0.5,
    "maxNewTokens": 10,
    "repetitionPenalty": 0.5,
    "maxLength": 1024,
    "prompt": "prompt",
    "model": "model",
    "url": "url",
    "parallel": 10,
    "logEveryQuery": 10,
    "api": "openai",
    "connectTimeout": 10,
    "readTimeout": 10,
    "zoneUuid": "zoneUuidExample",
    "name": "model-eval-service",
    "type": "VirtualMachine"
  },
  "systemTags": [],
  "userTags": []
}

上述示例中systemTagsuserTags字段可以省略。列出是为了表示body中可以包含这两个字段。

Curl示例
curl -H "Content-Type: application/json;charset=UTF-8" \
-H "Authorization: OAuth b86c9016b4f24953a9edefb53ca0678c" \
-X PUT -d '{"deployModelEvalService":{"taskType":"Performance","limits":10,"temperature":0.5,"topK":10,"topP":0.5,"maxNewTokens":10,"repetitionPenalty":0.5,"maxLength":1024,"prompt":"prompt","model":"model","url":"url","parallel":10,"logEveryQuery":10,"api":"openai","connectTimeout":10,"readTimeout":10,"zoneUuid":"zoneUuidExample","name":"model-eval-service","type":"VirtualMachine"}}' \
http://localhost:8080/zstack/v1/ai/model-services/eval/5634e459d1e13ff9abaf395a9517c100
参数列表
名字 类型 位置 描述 可选值 起始版本
limits Integer body(包含在deployModelEvalService结构中) 限制 5.1.8
temperature (可选) Float body(包含在deployModelEvalService结构中) 温度 5.1.8
topK (可选) Integer body(包含在deployModelEvalService结构中) Top K 5.1.8
topP (可选) Float body(包含在deployModelEvalService结构中) Top P 5.1.8
maxLength (可选) Integer body(包含在deployModelEvalService结构中) 最大长度 5.1.8
maxNewTokens (可选) Integer body(包含在deployModelEvalService结构中) 最大新令牌数 5.1.8
repetitionPenalty (可选) Float body(包含在deployModelEvalService结构中) 重复惩罚 5.1.8
uuid String url 资源的UUID,唯一标示该资源 5.1.8
description (可选) String body(包含在deployModelEvalService结构中) 资源的详细描述 5.1.8
modelUuid (可选) String body(包含在deployModelEvalService结构中) 模型UUID 5.1.8
zoneUuid String body(包含在deployModelEvalService结构中) 区域UUID 5.1.8
vmImageUuid (可选) String body(包含在deployModelEvalService结构中) 虚拟机镜像UUID 5.1.8
primaryStorageUuid (可选) String body(包含在deployModelEvalService结构中) 主存储UUID 5.1.8
datasetUuids (可选) List body(包含在deployModelEvalService结构中) 5.1.8
modelServiceGroupUuids (可选) List body(包含在deployModelEvalService结构中) 5.1.8
dockerImage (可选) String body(包含在deployModelEvalService结构中) Docker镜像 5.1.8
cpuNum (可选) Integer body(包含在deployModelEvalService结构中) 5.1.8
name String body(包含在deployModelEvalService结构中) 资源名称 5.1.8
type String body(包含在deployModelEvalService结构中)
  • VirtualMachine
  • Container
5.1.8
memorySize (可选) Long body(包含在deployModelEvalService结构中) 内存大小 5.1.8
l3NetworkUuids (可选) List body(包含在deployModelEvalService结构中) 5.1.8
systemTags (可选) List body 系统标签 5.1.8
userTags (可选) List body 用户标签 5.1.8
serviceBootUptime (可选) Integer body(包含在deployModelEvalService结构中) 5.1.20
taskType String body(包含在deployModelEvalService结构中) 任务类型
  • Performance
  • Capability
5.3.6
prompt (可选) String body(包含在deployModelEvalService结构中) 5.3.6
model (可选) String body(包含在deployModelEvalService结构中) 模型 5.3.6
url (可选) String body(包含在deployModelEvalService结构中) URL 5.3.6
parallel (可选) Integer body(包含在deployModelEvalService结构中) 并行 5.3.6
logEveryQuery (可选) Integer body(包含在deployModelEvalService结构中) 记录每个查询 5.3.6
api (可选) String body(包含在deployModelEvalService结构中) API
  • openai
  • dashscope
5.3.6
requestHeaders (可选) Map body(包含在deployModelEvalService结构中) 请求头 5.3.6
connectTimeout (可选) Integer body(包含在deployModelEvalService结构中) 连接超时 5.3.6
readTimeout (可选) Integer body(包含在deployModelEvalService结构中) 读取超时 5.3.6
environmentVariables (可选) Map body(包含在deployModelEvalService结构中) 5.3.20
startupParameters (可选) Map body(包含在deployModelEvalService结构中) 5.3.20
clusterUuid (可选) String body(包含在deployModelEvalService结构中) 集群UUID 5.3.22
rootDiskOfferingUuid (可选) String body(包含在deployModelEvalService结构中) 5.3.28
rootDiskSize (可选) Long body(包含在deployModelEvalService结构中) 5.3.28
resourceUuid (可选) String body(包含在deployModelEvalService结构中) 资源UUID 5.3.46
tagUuids (可选) List body(包含在deployModelEvalService结构中) 标签UUID列表 5.3.46
serviceLivez (可选) String body(包含在deployModelEvalService结构中) 5.4.6
serviceReadyz (可选) String body(包含在deployModelEvalService结构中) 5.4.6
requestCpuNum (可选) Integer body(包含在deployModelEvalService结构中) 请求CPU数量。 5.5.22
requestMemorySize (可选) Long body(包含在deployModelEvalService结构中) 请求内存大小。 5.5.22
projectUuid (可选) String body(包含在deployModelEvalService结构中) 项目UUID。 5.5.22

API返回

返回示例
{
  "inventory": {
    "uuid": "groupUuid",
    "name": "model-eval-service",
    "status": "Running"
  },
  "tasks": [
    {
      "uuid": "taskUuid",
      "name": "eval-task",
      "status": "Running"
    }
  ]
}
返回字段
名字 类型 描述 起始版本
success Boolean 操作是否成功 5.1.8
error ErrorCode 错误码,操作成功时该字段为null 5.1.8
inventory ModelEvalServiceInstanceGroupInventory 模型评测服务实例组清单 5.1.8
tasks List 模型评测任务清单列表 5.1.8
#inventory
名字 类型 描述 起始版本
uuid String 资源的UUID,唯一标识该资源 5.1.8
modelServiceUuid String 模型服务UUID 5.1.8
modelUuid String 模型UUID 5.1.8
instances List 模型服务实例列表 5.1.8
datasetRefInventories List 数据集引用列表 5.1.8
status String 实例组状态 5.1.8
modelServiceType String 模型服务类型 5.1.8
type String 实例组类型 5.1.8
name String 实例组名称 5.1.8
description String 实例组描述 5.1.8
yaml String YAML配置 5.1.8
supportMetrics List 支持的指标列表 5.1.8
exportPath String 导出路径 5.1.8
createDate Timestamp 创建时间 5.1.8
lastOpDate Timestamp 最后一次修改时间 5.1.8
limits Integer 限制数量 5.1.8
temperature Float 温度参数 5.1.8
topK Integer Top K参数 5.1.8
topP Float Top P参数 5.1.8
maxLength Integer 最大长度 5.1.8
maxNewTokens Integer 最大新生成Token数量 5.1.8
repetitionPenalty Float 重复惩罚参数 5.1.8
#tasks
名字 类型 描述 起始版本
uuid String 资源的UUID,唯一标识该资源 5.1.8
name String 任务名称 5.1.8
type String 任务类型 5.1.8
description String 任务描述 5.1.8
percentage Integer 任务进度百分比 5.1.8
opaque String 扩展信息 5.1.8
status String 任务状态 5.1.8
modelServiceGroupUuid String 模型服务实例组UUID 5.1.8
evaluatedServiceGroupUuid String 被评测服务实例组UUID 5.1.8
datasetUuid String 数据集UUID 5.1.8
limits Integer 限制数量 5.1.8
maxNewTokens Integer 最大新生成Token数量 5.1.8
topK Integer Top K参数 5.1.8
temperature Float 温度参数 5.1.8
topP Float Top P参数 5.1.8
prompt String 提示词 5.1.8
repetitionPenalty Float 重复惩罚参数 5.1.8
maxLength Integer 最大长度 5.1.8
model String 模型名称 5.1.8
url String 服务URL 5.1.8
parallel Integer 并发数 5.1.8
logEveryQuery Integer 查询日志间隔 5.1.8
api String API类型 5.1.8
requestHeaders Map 请求头 5.1.8
connectTimeout Integer 连接超时时间 5.1.8
readTimeout Integer 读取超时时间 5.1.8
totalScore Double 总分 5.1.8
endTime Timestamp 结束时间 5.1.8
createDate Timestamp 创建时间 5.1.8
lastOpDate Timestamp 最后一次修改时间 5.1.8

SDK示例

Java SDK
DeployModelEvalServiceAction action = new DeployModelEvalServiceAction();
action.taskType = "Performance";
action.limits = 10;
action.temperature = 0.5;
action.topK = 10;
action.topP = 0.5;
action.maxNewTokens = 10;
action.repetitionPenalty = 0.5;
action.maxLength = 1024;
action.prompt = "prompt";
action.model = "model";
action.url = "url";
action.parallel = 10;
action.logEveryQuery = 10;
action.api = "openai";
action.connectTimeout = 10;
action.readTimeout = 10;
action.zoneUuid = "zoneUuidExample";
action.name = "model-eval-service";
action.type = "VirtualMachine";
action.uuid = "5634e459d1e13ff9abaf395a9517c100";
action.sessionId = "b86c9016b4f24953a9edefb53ca0678c";
DeployModelEvalServiceAction.Result res = action.call();
Python SDK
DeployModelEvalServiceAction action = DeployModelEvalServiceAction()
action.taskType = "Performance"
action.limits = 10
action.temperature = 0.5
action.topK = 10
action.topP = 0.5
action.maxNewTokens = 10
action.repetitionPenalty = 0.5
action.maxLength = 1024
action.prompt = "prompt"
action.model = "model"
action.url = "url"
action.parallel = 10
action.logEveryQuery = 10
action.api = "openai"
action.connectTimeout = 10
action.readTimeout = 10
action.zoneUuid = "zoneUuidExample"
action.name = "model-eval-service"
action.type = "VirtualMachine"
action.uuid = "5634e459d1e13ff9abaf395a9517c100"
action.sessionId = "b86c9016b4f24953a9edefb53ca0678c"
DeployModelEvalServiceAction.Result res = action.call()