
Ultimate access to all questions.
Your data science team is working on optimizing a PyTorch model for image classification, leveraging a pre-trained ResNet model. The team aims to achieve the best performance through hyperparameter tuning. The project has constraints including the need for scalability, cost-effectiveness, and minimal changes to the existing PyTorch framework. Which of the following approaches should you take to meet these requirements? (Choose two options if E is available)
A
Transform the model into a TensorFlow model and execute a hyperparameter tuning job on AI Platform, despite the additional conversion overhead.
B
Switch the model to a Keras model and utilize Keras Tuner for hyperparameter optimization, considering the ease of use but potential limitations with PyTorch.
C
Execute a hyperparameter tuning job on AI Platform with custom containers, allowing for direct control over the PyTorch environment without framework conversion.
D
Deploy a Kubeflow Pipelines instance and conduct hyperparameter tuning using Katib, which supports PyTorch natively but requires additional infrastructure setup.
E
Utilize both AI Platform with custom containers for immediate needs and Kubeflow Pipelines for long-term scalability and flexibility.