查询超参搜索所有trial的结果 - ShowAutoSearchTrials

功能介绍

查询超参搜索所有trial的结果接口用于获取指定超参搜索任务中所有trial的详细结果。

该接口适用于以下场景：当用户需要查看超参搜索任务中所有trial的性能指标和状态信息时，可以通过此接口获取详细结果。使用该接口的前提条件是超参搜索任务已存在且用户具有相应的权限。查询操作完成后，接口将返回所有trial的详细结果信息。若超参搜索任务不存在或用户无权限操作，接口将返回相应的错误信息。

调试

您可以在API Explorer中调试该接口，支持自动认证鉴权。API Explorer可以自动生成SDK代码示例，并提供SDK代码示例调试功能。获取该接口关联的CLI示例： hcloud ModelArts ShowAutoSearchTrials。

授权信息

账号具备所有API的调用权限，如果使用账号下的IAM用户调用当前API，该IAM用户需具备调用API所需的权限。

• 如果使用角色与策略授权，具体权限要求请参见权限和授权项。
• 如果使用身份策略授权，需具备如下身份策略权限。

  授权项	访问级别	资源类型（*为必须）	条件键	别名	依赖的授权项
  modelarts:trainJob:get	Read	trainJob *	g:ResourceTag/<tag-key>	-	-
  		-	modelarts:poolType modelarts:poolId

URI

GET /v2/{project_id}/training-jobs/{training_job_id}/autosearch-trials

请求参数

无

响应参数

状态码：200

请求示例

如以下查询training_job_id为5b60a667-1438-4eb5-9705-85b860e623dc的作业的所有trial的信息。

GET https://endpoint/v2/{project_id}/training-jobs/5b60a667-1438-4eb5-9705-85b860e623dc/autosearch-trials

响应示例

状态码：200

ok

{
  "total" : 8,
  "count" : 8,
  "limit" : 50,
  "offset" : 0,
  "group_by" : "",
  "items" : {
    "header" : [ "", "done", "pid", "config", "trial_id", "training_iteration", "time_total_s", "worker_index", "reward_attr", "status", "acc", "loss", "best_reward" ],
    "data" : [ [ "0", "True", "314", {
      "batch_size" : 32,
      "learning_rate" : 0.05512301741232006,
      "trial_index" : 0,
      "param/batch_size" : 32,
      "param/learning_rate" : 0.05512301741232006
    }, "ae544174", "2", "19.477163314819336", "", "0.0625", "TERMINATED", "0.0625", "tensor(0.0754, device='cuda:0', requires_grad=True)", "0.0625" ], [ "1", "True", "315", {
      "batch_size" : 32,
      "learning_rate" : 0.0785570955603036,
      "trial_index" : 1,
      "param/batch_size" : 32,
      "param/learning_rate" : 0.0785570955603036
    }, "ae548666", "2", "3.601897954940796", "", "0.0", "TERMINATED", "0.0", "tensor(0.0760, device='cuda:0', requires_grad=True)", "0.0" ], [ "2", "True", "312", {
      "batch_size" : 16,
      "learning_rate" : 0.04015387428829642,
      "trial_index" : 2,
      "param/batch_size" : 16,
      "param/learning_rate" : 0.04015387428829642
    }, "ae54c0ea", "2", "3.5978384017944336", "", "0.1875", "TERMINATED", "0.1875", "tensor(0.1469, device='cuda:0', requires_grad=True)", "0.1875" ], [ "3", "True", "313", {
      "batch_size" : 32,
      "learning_rate" : 0.0340820322164706,
      "trial_index" : 3,
      "param/batch_size" : 32,
      "param/learning_rate" : 0.0340820322164706
    }, "ae5503c0", "2", "3.641200304031372", "", "0.25", "TERMINATED", "0.25", "tensor(0.0716, device='cuda:0', requires_grad=True)", "0.25" ], [ "4", "True", "470", {
      "batch_size" : 32,
      "learning_rate" : 0.03656488928171769,
      "trial_index" : 4,
      "param/batch_size" : 32,
      "param/learning_rate" : 0.03656488928171769
    }, "bef46590", "2", "3.6120550632476807", "", "0.09375", "TERMINATED", "0.09375", "tensor(0.0740, device='cuda:0', requires_grad=True)", "0.09375" ], [ "5", "True", "499", {
      "batch_size" : 32,
      "learning_rate" : 0.008413169003970163,
      "trial_index" : 5,
      "param/batch_size" : 32,
      "param/learning_rate" : 0.008413169003970163
    }, "bef578f4", "2", "3.6379287242889404", "", "0.1875", "TERMINATED", "0.1875", "tensor(0.0723, device='cuda:0', requires_grad=True)", "0.1875" ], [ "6", "True", "528", {
      "batch_size" : 64,
      "learning_rate" : 0.06297447200613912,
      "trial_index" : 6,
      "param/batch_size" : 64,
      "param/learning_rate" : 0.06297447200613912
    }, "bef5c584", "2", "3.711118221282959", "", "0.046875", "TERMINATED", "0.046875", "tensor(0.0381, device='cuda:0', requires_grad=True)", "0.046875" ], [ "7", "True", "557", {
      "batch_size" : 32,
      "learning_rate" : 0.04426479392014276,
      "trial_index" : 7,
      "param/batch_size" : 32,
      "param/learning_rate" : 0.04426479392014276
    }, "bef60684", "2", "3.6971280574798584", "", "0.0625", "TERMINATED", "0.0625", "tensor(0.0778, device='cuda:0', requires_grad=True)", "0.0625" ] ]
  }
}

状态码

错误码

请参见错误码。