MLS-C01 · Question #27
A company's Machine Learning Specialist needs to improve the training speed of a time-series forecasting model using TensorFlow. The training is currently implemented on a single-GPU machine and…
The correct answer is B. Change the TensorFlow code to implement a Horovod distributed framework supported by. Horovod is an open-source distributed deep learning framework that integrates directly with TensorFlow (and other frameworks) with minimal code changes - typically just a few lines to wrap the optimizer and broadcast initial variables. Amazon SageMaker natively supports Horovod…
Question
A company's Machine Learning Specialist needs to improve the training speed of a time-series forecasting model using TensorFlow. The training is currently implemented on a single-GPU machine and takes approximately 23 hours to complete. The training needs to be run daily. The model accuracy is acceptable, but the company anticipates a continuous increase in the size of the training data and a need to update the model on an hourly, rather than a daily, basis. The company also wants to minimize coding effort and infrastructure changes. What should the Machine Learning Specialist do to the training solution to allow it to scale for future demand?
Options
- ADo not change the TensorFlow code. Change the machine to one with a more powerful GPU to
- BChange the TensorFlow code to implement a Horovod distributed framework supported by
- CSwitch to using a built-in AWS SageMaker DeepAR model. Parallelize the training to as many
- DMove the training to Amazon EMR and distribute the workload to as many machines as needed to
How the community answered
(24 responses)- A21% (5)
- B67% (16)
- C4% (1)
- D8% (2)
Explanation
Horovod is an open-source distributed deep learning framework that integrates directly with TensorFlow (and other frameworks) with minimal code changes - typically just a few lines to wrap the optimizer and broadcast initial variables. Amazon SageMaker natively supports Horovod for distributed training across multiple GPUs and instances, allowing the workload to scale horizontally as data grows. This meets the requirements of minimal code change and minimal infrastructure change. Upgrading to a single more-powerful GPU (A) hits hardware limits and won't scale to hourly training as data grows indefinitely. Switching to DeepAR (C) is a model change, not a training optimization. Moving to EMR (D) represents a significant infrastructure shift and is not optimized for deep learning workloads.
Topics
Community Discussion
No community discussion yet for this question.