0

有关如何指定 Google 预测存储桶的任何经验。我继续通过单击按钮创建存储桶在 Storage --> Cloud Storage --> Browser 中创建了一个存储桶。我将存储桶命名为唯一名称，例如“UNIQUE001”。我根据他们的规范上传了数据，一个 CSV 文件，如果字符串被引用。我想训练模型，所以想运行 Prediction API v1.6 > prediction.trainedmodels.insert。这是我的发帖请求

POST https://www.googleapis.com/prediction/v1.6/projects/XXXXXXXXX/trainedmodels?fields=created%2Cid%2Ckind%2CmodelInfo%2CmodelType%2CselfLink%2CstorageDataLocation%2CstoragePMMLLocation%2CstoragePMMLModelLocation%2CtrainingComplete%2CtrainingStatus

    {
     "id": "CodePrediction",
     "storageDataLocation": "dataset.csv",
     "modelType": "CLASSIFICATION"
    }

我收到错误“未找到训练数据”的响应 400 OK

显示标题 -

{
 "error": {
  "errors": [
   {
    "domain": "global",
    "reason": "invalid",
    "message": "Training data file not found."
   }
  ],
  "code": 400,
  "message": "Training data file not found."
 }
}

不确定，我猜如何指定 storageDataLocation。我试过 1) gs://UNIQUE001 2) ProjectNumber/UNIQUE001

这是我给出 ProjectNumber/bucketName 时的错误，

Request
POST https://www.googleapis.com/prediction/v1.6/projects/xxxx/trainedmodels?fields=created%2Cid%2Ckind%2CmodelInfo%2CmodelType%2CselfLink%2CstorageDataLocation%2CstoragePMMLLocation%2CstoragePMMLModelLocation%2CtrainingComplete%2CtrainingStatus

{
 "id": "CodePrediction",
 "storageDataLocation": "xxxxxxx/UNIQUE001",
 "modelType": "CLASSIFICATION"
}

错误

400 好

显示标题 -

{
 "error": {
  "errors": [
   {
    "domain": "global",
    "reason": "invalid",
    "message": "Training data file is empty.",
    "locationType": "other",
    "location": "id"
   }
  ],
  "code": 400,
  "message": "Training data file is empty."
 }
}

数据集不为空。它有百万行

4

1 回答 1

0

storageDataLocation 值为 bucketname/filename

于 2015-10-05T18:00:28.640 回答