Google Professional Machine Learning Engineer Question 183
Select 2Google Cloud PlatformYou are tasked with building a machine learning solution for a client that needs to classify large volumes of customer support emails into predefined categories such as 'Billing', 'Technical Support', and 'General Inquiry'. The client wants a quick solution without significant investment in training a custom model from scratch. You decide to use Google Cloud's Model Garden for this task. Which of the following steps should you take to efficiently leverage a pre-trained model from Model Garden to meet the client's requirements?
- A
Search Model Garden for a pre-trained foundational model specialized in text classification and fine-tune it with the client's labeled data.
- B
Deploy a pre-trained model from Model Garden as-is, without any modifications, and use it to process the client's emails.
- C
Export a foundational image classification model from Model Garden and adapt it for text classification using transfer learning.
- D
Use Google Vertex AI to deploy the fine-tuned model from Model Garden into production and integrate it with the client’s email processing pipeline.
- E
Train a custom model from scratch using TensorFlow and ignore the pre-trained models available in Model Garden.
Show answer and explanation
Correct answers: A, D
Explanation
To efficiently leverage Google Cloud's Model Garden, you should search for a pre-trained foundational model that aligns with the task at hand (e.g., text classification). Fine-tuning the model with the client’s labeled data ensures that the model is optimized for the specific use case. Once fine-tuned, deploying the model using Vertex AI allows you to integrate it seamlessly into the client’s operational pipeline. This approach saves time and resources compared to training a custom model from scratch or repurposing unrelated models.
- A. Correct.
Correct. Searching for a pre-trained model specialized in text classification and fine-tuning it with the client’s labeled data is an efficient approach to adapt the model to the client's specific categories.
- B. Incorrect.
Incorrect. While deploying a pre-trained model as-is might work in some cases, it will likely not provide the best accuracy for the client's specific categories without fine-tuning.
- C. Incorrect.
Incorrect. Foundational image classification models are not suitable for text classification tasks. Adapting such a model would require significant effort and is not recommended.
- D. Correct.
Correct. Vertex AI is the recommended service for deploying a fine-tuned model into production and integrating it into the client’s workflow.
- E. Incorrect.
Incorrect. Training a custom model from scratch is time-consuming and unnecessary when pre-trained models from Model Garden exist for this use case.