Amazon
MLS-C01 · Question #253
A data scientist is training a large PyTorch model by using Amazon SageMaker. It takes 10 hours on average to train the model on GPU instances. The data scientist suspects that training is not…
The correct answer is C. Use the SageMaker Debugger vanishing_gradient and LowGPUUtilization built-in rules to detect. https://docs.aws.amazon.com/sagemaker/latest/dg/debugger-built-in-rules.html
Machine Learning Implementation and Operations
Question
A data scientist is training a large PyTorch model by using Amazon SageMaker. It takes 10 hours on average to train the model on GPU instances. The data scientist suspects that training is not converging and that resource utilization is not optimal. What should the data scientist do to identify and address training issues with the LEAST development effort?
Options
- AUse CPU utilization metrics that are captured in Amazon CloudWatch. Configure a CloudWatch
- BUse high-resolution custom metrics that are captured in Amazon CloudWatch. Configure an AWS
- CUse the SageMaker Debugger vanishing_gradient and LowGPUUtilization built-in rules to detect
- DUse the SageMaker Debugger confusion and feature_importance_overweight built-in rules to
How the community answered
(46 responses)- A2% (1)
- B4% (2)
- C83% (38)
- D11% (5)
Explanation
https://docs.aws.amazon.com/sagemaker/latest/dg/debugger-built-in-rules.html
Topics
#SageMaker Debugger#Model Training Optimization#Resource Utilization Monitoring#Convergence Issues
Community Discussion
No community discussion yet for this question.