Google Professional Machine Learning Engineer Question 38
Select 3Google Cloud PlatformYour company is building a customer support chatbot to process user queries efficiently. The team has decided to use a pre-trained text classification model from Google Cloud Model Garden to classify incoming queries into categories like 'Billing', 'Technical Support', and 'General Inquiry'. After selecting the model, what steps should you take to integrate it into your application effectively?
- A
Download the model locally and deploy it on your on-premises servers.
- B
Use the Vertex AI Prediction service to deploy the pre-trained model directly from Model Garden.
- C
Fine-tune the pre-trained model on your custom dataset using Vertex AI Workbench before deploying.
- D
Retrieve the model's serving endpoint from the Vertex AI dashboard and call it from your application.
- E
Manually write custom TensorFlow code to serve the model on Google Cloud.
Show answer and explanation
Correct answers: B, C, D
Explanation
To integrate a pre-trained model from Google Cloud Model Garden, you can deploy it directly to the Vertex AI Prediction service. Fine-tuning the model can improve its performance for your specific use case, and the serving endpoint can be used to communicate with the deployed model. Google Cloud's managed services eliminate the need for manual deployment or custom TensorFlow serving code, simplifying the process.
- A. Incorrect.
Downloading the model for on-premises deployment is not necessary when using Google Cloud's Model Garden. The models are designed to run on Google's managed services.
- B. Correct.
Using the Vertex AI Prediction service is the recommended way to deploy pre-trained models from Model Garden. This allows you to use a fully managed environment for serving.
- C. Correct.
Fine-tuning the pre-trained model on your specific dataset helps improve its accuracy for your specific use case. This step is optional but often beneficial.
- D. Correct.
Retrieving the model's serving endpoint from the Vertex AI dashboard is the correct way to integrate the model into your application. This allows your application to send requests directly to the deployed model.
- E. Incorrect.
Manually writing custom TensorFlow code to serve the model is unnecessary. Google Cloud provides managed services like Vertex AI to handle model serving.