Google Professional Machine Learning Engineer exam dumps

Google Professional Machine Learning Engineer practice question 307 of 522

Professional Machine Learning Engineer. Professional level, Google Cloud. Free question with the correct answer and a full explanation.

Google Professional Machine Learning Engineer Question 307

Select 3Google Cloud Platform

You are working on deploying a machine learning model on Google Cloud. Your team has trained the model using PyTorch, but the serving platform your organization uses requires TensorFlow SavedModel format for deployment. Which of the following steps would allow you to successfully deploy your PyTorch model on the serving platform?

  1. A

    Convert the PyTorch model to ONNX format and then convert it to TensorFlow SavedModel format.

  2. B

    Deploy the PyTorch model directly using AI Platform Prediction without any conversion.

  3. C

    Export the PyTorch model as TorchScript and use a custom prediction routine to serve it.

  4. D

    Use TensorFlow's tf.keras.models.load_model() to directly load the PyTorch model.

  5. E

    Convert the PyTorch model to TensorFlow SavedModel format using an appropriate library like ONNX or a custom script.

Show answer and explanation

Correct answers: A, C, E

Explanation

When serving models trained in PyTorch on a platform that requires TensorFlow SavedModel format, you can either convert the PyTorch model to TensorFlow SavedModel format (e.g., via ONNX or a custom script) or use an alternative like exporting the model as TorchScript and serving it with a custom routine. Direct deployment without conversion or loading a PyTorch model in TensorFlow's API is not possible due to incompatibilities between the frameworks.

  • A. Correct.

    Correct. Converting the PyTorch model to ONNX format and then to TensorFlow SavedModel format is a valid approach because ONNX serves as an intermediate representation that is interoperable with various frameworks.

  • B. Incorrect.

    Incorrect. While AI Platform Prediction allows deployment of models, it does not natively support PyTorch models without a custom container or conversion.

  • C. Correct.

    Correct. Exporting the model as TorchScript and using a custom prediction routine is an alternative approach to serving PyTorch models on AI Platform Prediction.

  • D. Incorrect.

    Incorrect. TensorFlow's tf.keras.models.load_model() cannot load PyTorch models directly as they are fundamentally different frameworks.

  • E. Correct.

    Correct. Using appropriate tools or libraries like ONNX or a custom script to convert a PyTorch model to TensorFlow SavedModel format ensures compatibility with TensorFlow-based serving platforms.

Timed practice exam

Take a Google Professional Machine Learning Engineer practice test under exam conditions

60 questions in 120 minutes, drawn from this bank, with a score report and a per-question review when you finish.

Start timed exam

More Google Professional Machine Learning Engineer practice questions

  1. 308.You are tasked with deploying a machine learning model built using PyTorch for a production...
  2. 309.Your team is building a machine learning platform on Google Cloud for deploying multiple models and...
  3. 310.You are a machine learning engineer managing a model registry for your organization using Vertex...
  4. 311.You are working on a machine learning project where multiple teams collaborate on developing and...
  5. 312.Your organization has implemented several machine learning models that are used across different...
Back to the Google Professional Machine Learning Engineer exam dumps