Professional Machine Learning Engineer

Question

You created a custom ML model with scikit-learn. Its training takes longer than anticipated. You decide to move the model to Vertex AI Training and want to improve training time. What should you try first?

Explanation

scikit-learn training commonly depends on NumPy and SciPy numerical operations, so using their optimized internal/vectorized routines on a Deep Learning VM is the appropriate first performance improvement without rewriting the model. Deep Learning VM images are optimized for ML workloads and include NumPy, SciPy, and scikit-learn. Distributed training and GPU acceleration require workload-specific support and can introduce additional overhead, while migrating to TensorFlow requires an unnecessary model rewrite.

Learn more

Introduction to Deep Learning VM

Professional Machine Learning Engineer

AstroTutor

Notes

Readiness

QuestionQ46

Community Discussion