如何使用HTTP请求对GCP AI平台预测进行身份验证

问题描述

通过遵循教程.现在，我想通过HTTP请求对此模型进行预测.

I have deployed a model to AI Platform Predictions successfully by following this tutorial. Now, I want to make predictions with this model via HTTP requests.

按照@Ismail的建议，我遵循了方法:projects.predict 文档.

As suggested by @Ismail, I followed Method: projects.predict documentation.

我正在向 https://ml.googleapis.com/v1/projects/$ {PROJECT_ID}/models/$ {MODEL_NAME}:predict 发送POST请求，其正文如下:

I'm sending a POST request to https://ml.googleapis.com/v1/projects/${PROJECT_ID}/models/${MODEL_NAME}:predict with the following body:

{
  "instances": [
    [51.0, 17.0, 6.0, 1.0, 1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1.0, 0.0, 0.0, 0.0, 0.0, 1.0, 0.0, 1.0, 0.0, 0.0]
  ]
}

数字是模型的输入.

我收到以下答复:

{
    "error": {
        "code": 401,
        "message": "Request is missing required authentication credential. Expected OAuth 2 access token, login cookie or other valid authentication credential. See https://developers.google.com/identity/sign-in/web/devconsole-project.",
        "status": "UNAUTHENTICATED"
    }
}

我使用RestMan chrome扩展程序发送了POST请求.我希望能够通过RestMan获得预测.

I sent the POST request using the RestMan chrome extension. I would like to be able to get a prediction with RestMan.

我的问题是:如何验证我的HTTP POST请求?我应该在请求的标题中发送一些信息吗?

My question is: how can I authenticate my HTTP POST request? Should I send some information in the request's header?

推荐答案

访问Google API时，需要在标头中添加访问令牌.使用CLI，您可以像这样生成它:

When you access to Google API, you need to add an access token in your header. With the CLI you can generate it like this:

gcloud auth print-access-token

当然，当前的凭据需要在AI平台预测服务上得到授权.

如果执行卷曲，则可以在linux(和云外壳)上执行

If you perform a curl, you can do this on linux (and cloud shell)

curl -X POST -d @bodyfile.json -H "Authorization: Bearer $(gcloud auth print-access-token)" \
  https://ml.googleapis.com/v1/projects/${PROJECT_ID}/models/${MODEL_NAME}:predic

---

如果您使用RestMan，我不确定它是否能够为您生成访问令牌.解决方案是公开您的模型.不建议这样做，但是要在有限的时间内(仅用于测试)，为什么不这样做！

---

If you use RestMan, I'm not sure it is capable to generate an access token for you. The solution is to make your model public. It's not recommended, but for a limited period of time, and for test only, why not!

为此，请为模型中的角色 roles/ml.modelUser 授予 allUsers .

For this, grant allUsers with the role roles/ml.modelUser on your model.

编辑1

我没有找到如何在带有控制台的模型上执行此操作.我以前没做过，但是有一些更新...很奇怪.我通过gcloud CLI实现了这一点

I didn't find how to do this on a model with the console. I did t before but there were some updates... Very strange. I achieve this with the gcloud CLI

gcloud ai-platform models add-iam-policy-binding --region=us-central1 --member=allUsers --role=roles/ml.modelUser test

这篇关于如何使用HTTP请求对GCP AI平台预测进行身份验证的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持！