You are developing ML models with AI Platform for image segmentation on CT scans.
You frequently update your model architectures based on the newest available research papers, and have to rerun training on the same dataset to benchmark their performance.
You want to minimize computation costs and manual intervention while having version control for your code.
What should you do?
Click on the arrows to vote for the correct answer
A. B. C. D.B.
https://cloud.google.com/ai-platform/training/docs/training-jobsIn this scenario, you want to minimize computation costs and manual intervention while having version control for your code when updating model architectures for image segmentation on CT scans using AI Platform.
Option A: Using Cloud Functions to identify changes to your code in Cloud Storage and trigger a retraining job.
Cloud Functions is a serverless platform that allows you to run code in response to events, such as changes to objects in Cloud Storage. This option can be a good fit if you want to automate the training process when you update your code, and you prefer a serverless approach. However, Cloud Functions is not designed to handle long-running jobs like training a machine learning model. Thus, this option is not recommended for training ML models.
Option B: Using the gcloud command-line tool to submit training jobs on AI Platform when you update your code.
This option involves manually submitting training jobs on AI Platform using the gcloud command-line tool when you update your code. While it provides version control for your code and flexibility in terms of parameter tuning, it requires manual intervention to trigger training jobs, which can be time-consuming and error-prone. Furthermore, the lack of automation can lead to inconsistent results between training runs.
Option C: Using Cloud Build linked with Cloud Source Repositories to trigger retraining when new code is pushed to the repository.
Cloud Build is a fully-managed build service that allows you to build, test, and deploy code to various environments, including AI Platform. Cloud Source Repositories is a Git-based source code repository that provides version control for your code. This option enables you to automate the training process by triggering training jobs when new code is pushed to the repository. It also provides version control and can help reduce manual intervention. This option is recommended for its automation, version control, and reliability.
Option D: Creating an automated workflow in Cloud Composer that runs daily and looks for changes in code in Cloud Storage using a sensor.
Cloud Composer is a fully-managed workflow orchestration service based on Apache Airflow. It enables you to create, schedule, and monitor workflows that integrate with various services, including AI Platform. This option involves creating a workflow that runs daily and checks for changes in code in Cloud Storage using a sensor. When a change is detected, the workflow triggers a training job on AI Platform. This option provides automation, version control, and scheduling flexibility, making it an excellent choice for ML engineers who want to minimize manual intervention and computation costs.
In conclusion, the recommended option for minimizing computation costs and manual intervention while having version control for your code when updating model architectures for image segmentation on CT scans using AI Platform is Option C: Using Cloud Build linked with Cloud Source Repositories to trigger retraining when new code is pushed to the repository. This option provides automation, version control, and reliability. However, if you require more flexibility in scheduling, you can consider Option D: Creating an automated workflow in Cloud Composer that runs daily and looks for changes in code in Cloud Storage using a sensor.