AWS Certified Machine Learning Engineer Associate · Free Practice Question Medium
Question 12
You are training a machine learning model using PyTorch on Amazon SageMaker and want to ensure that the system resources, such as CPU and GPU, are being effectively utilized. You plan to use SageMaker Profiler to track these metrics and improve the efficiency of your training job. Which of the following steps should you follow to set up and use SageMaker Profiler correctly? (Select Two)
-
A
Import the necessary SageMaker Profiler modules and add start and stop profiling commands in your PyTorch training script.
-
B
Use SageMaker Neo to compile your model for optimized CPU and GPU performance after training is completed.
-
C
Configure the SageMaker Profiler to automatically tune your hyperparameters, such as learning rate, based on the resource utilization data.
-
D
Define the profiler configuration in your SageMaker Estimator, specifying the duration of the profiling in seconds for the CPU and GPU.
-
E
Manually enable automatic SageMaker Profiler optimization in the SageMaker console to reduce memory overhead during training.
Reveal correct answers
Correct answers: A, D
A.
To track and monitor system resource usage, the SageMaker Profiler modules need to be imported, and start/stop profiling commands must be incorporated into the training script. This ensures that the profiler is actively capturing performance data during training.
B.
SageMaker Neo is used for compiling models post-training to optimize their performance for specific deployment hardware environments. It is not used during the training process to track system resources.
C.
SageMaker Profiler monitors system resource utilization but does not automatically tune hyperparameters such as learning rate. Hyperparameter tuning is a separate process in SageMaker, managed through SageMaker's automatic model tuning capabilities.
D.
You need to configure the SageMaker Estimator with profiling settings to capture CPU and GPU utilization data at specific intervals during the training job. This allows for effective monitoring and optimization of resource usage.
E.
There is no feature in the SageMaker console to automatically reduce memory overhead using the SageMaker Profiler. Resource monitoring requires manual setup in the training script or Estimator configuration.
Discussion
Think the marked answer is wrong, or have a better explanation? Share it below — comments appear after review.
