表单识别器“缺少训练数据：在给定路径找不到任何训练数据。”

Question

I am using new form recognizer studio ( https://formrecognizer.appliedai.azure.com/studio/customform/projects ) to label files using SDK Version="4.0.0-beta.2". Code to train model as below:

                var buildModelOptions = new BuildModelOptions
                {
                    Prefix = docType.SubFolderName,
                    ModelDescription = trainedModel.ModelDescription
                };
                BuildModelOperation operation = await _trainingClient
                    .StartBuildModelAsync(new Uri(_trainingDataSASUrl), trainedModel.ModelName, buildModelOptions);
                Response<DocumentModel> operationResponse = await operation.WaitForCompletionAsync();
                DocumentModel model = operationResponse.Value;

Keep getting error like below on method StartBuildModelAsync():

Invalid request.
Status: 400 (Bad Request)
ErrorCode: InvalidRequest

Content:
{
  "error": {
    "code": "InvalidRequest",
    "message": "Invalid request.",
    "innererror": {
      "code": "TrainingContentMissing",
      "message": "Training data is missing: Could not find any training data at the given path."
    }
  }
}

In BuildModelOptions.Prefix, I am passing value as "cfs/Lease1015", which is path from blob container, blob containers SAS url is in "_trainingDataSASUrl". In blob training files exist under "cfs/Lease1015" and prefix is specified in correct casing.

Answer 1

Please look at github bug reported here https://github.com/Azure/azure-sdk-for-net/issues/26081 for more information.

It worked after adding '/' to prefix attribute.

var buildModelOptions = new BuildModelOptions
{
    Prefix = docType.SubFolderName + '/', // append a slash to the subfolder name
    ModelDescription = trainedModel.ModelDescription
};