MLS-C01 Question 156
Select 4You are training a machine learning model using Amazon SageMaker, and the dataset is extremely large, causing training jobs to take a significant amount of time. Which of the following actions can you take to optimize the training process while ensuring model accuracy?
- A
Use a SageMaker built-in algorithm that supports distributed training.
- B
Reduce the size of the dataset by randomly removing rows to decrease training time.
- C
Enable SageMaker Pipe mode to stream data directly from Amazon S3 to the training instance.
- D
Use an appropriate instance type with GPU acceleration for faster training.
- E
Preprocess the dataset to remove unnecessary features and reduce dimensionality.
Show answer and explanation
Correct answers: A, C, D, E
Explanation
Optimizing model training in Amazon SageMaker for large datasets involves using the right tools and techniques to improve efficiency without compromising accuracy. Distributed training, Pipe mode, GPU acceleration, and feature preprocessing all contribute to faster and more efficient training. However, simply reducing the dataset size by randomly removing rows is not a recommended practice, as it can lead to loss of critical information and negatively impact the model's performance.
- A. Correct.
Using a SageMaker built-in algorithm that supports distributed training allows you to train large datasets more efficiently by distributing the workload across multiple instances.
- B. Incorrect.
Randomly removing rows from the dataset can harm the overall accuracy of the model, as important data points might be lost. This is not a recommended approach.
- C. Correct.
Enabling SageMaker Pipe mode allows data to be streamed directly from Amazon S3 to the training instance, reducing the input/output (I/O) overhead and improving training efficiency.
- D. Correct.
Using an instance type with GPU acceleration is a proven way to speed up training for models that can leverage parallel processing, such as deep learning models.
- E. Correct.
Preprocessing the dataset to remove unnecessary features and reduce dimensionality can improve model efficiency and decrease training time without sacrificing accuracy.