共用方式為


Custom Models - Train

Create 並定型自定義模型。 要求必須包含來源參數,該參數可能是外部可存取的 Azure 記憶體 Blob 容器 URI (最好是共用存取簽章 URI) 或本機掛接磁碟驅動器中數據資料夾的有效路徑。 指定本機路徑時,它們必須遵循Linux/Unix路徑格式,而且是根目錄為輸入掛接組態設定值的絕對路徑,例如,如果 '{Mounts:Input}' 組態設定值為 '/input',則有效的來源路徑會是 '/input/contosodataset'。 所有要定型的數據應該都位於源資料夾或子資料夾底下。 模型是使用下列內容類型的檔來定型 - 'application/pdf'、'image/jpeg'、'image/png'、'image/tiff' 或 'image/bmp'。 忽略其他類型的內容。

POST {endpoint}/formrecognizer/v2.1/custom/models

URI 參數

名稱 位於 必要 類型 Description
endpoint
path True

string

支援的認知服務端點 (通訊協定和主機名,例如: https://westus2.api.cognitive.microsoft.com) 。

要求標頭

名稱 必要 類型 Description
Ocp-Apim-Subscription-Key True

string

要求本文

名稱 必要 類型 Description
source True

string

包含定型檔的來源路徑。

modelName

string

選擇性使用者定義模型名稱 (長度上限:1024) 。

sourceFilter

TrainSourceFilter

篩選以套用至來源路徑中的檔以進行定型。

useLabelFile

boolean

使用標籤檔案來定型模型。

回應

名稱 類型 Description
201 Created

要求已順利排入佇列。

標題

Location: string

Other Status Codes

ErrorResponse

回應實體隨附的非成功回應,其中包含有關錯誤的其他詳細數據。

安全性

Ocp-Apim-Subscription-Key

類型: apiKey
位於: header

範例

Train custom model
Train custom model with subfolder filter options

Train custom model

範例要求

POST {endpoint}/formrecognizer/v2.1/custom/models


{
  "source": "{azure_blob_endpoint}/input/data1?sasToken"
}

範例回覆

Location: {endpoint}/formrecognizer/v2.1/custom/models/f973e3c1-1148-43bb-bea8-49d0603ab3a8

Train custom model with subfolder filter options

範例要求

POST {endpoint}/formrecognizer/v2.1/custom/models


{
  "source": "{azure_blob_endpoint}/input/data1?sasToken",
  "sourceFilter": {
    "prefix": "",
    "includeSubFolders": false
  },
  "useLabelFile": false
}

範例回覆

Location: {endpoint}/formrecognizer/v2.1/custom/models/f973e3c1-1148-43bb-bea8-49d0603ab3a8

定義

名稱 Description
ErrorInformation
ErrorResponse
TrainRequest

要求參數來定型新的自定義模型。

TrainSourceFilter

篩選以套用至來源路徑中的檔以進行定型。

ErrorInformation

名稱 類型 Description
code

string

message

string

ErrorResponse

名稱 類型 Description
error

ErrorInformation

TrainRequest

要求參數來定型新的自定義模型。

名稱 類型 預設值 Description
modelName

string

選擇性使用者定義模型名稱 (長度上限:1024) 。

source

string

包含定型檔的來源路徑。

sourceFilter

TrainSourceFilter

篩選以套用至來源路徑中的檔以進行定型。

useLabelFile

boolean

False

使用標籤檔案來定型模型。

TrainSourceFilter

篩選以套用至來源路徑中的檔以進行定型。

名稱 類型 預設值 Description
includeSubFolders

boolean

False

旗標,指出搜尋要前置處理的內容時,是否也需要包含前置詞資料夾集合內的子資料夾。

prefix

string

區分大小寫的前置詞字串,可篩選來源路徑中的檔以進行定型。 例如,使用 Azure 記憶體 Blob Uri 時,請使用前置詞來限制子資料夾進行定型。