Exam AWS Certified Machine Learning - Specialty topic 1 question 9 discussion

Answer is A. The answer to this question is about Pipe mode from S3. The only options are A and C. As AWS Glue cannot be use to create models which is option C. The correct answer is A

Answer is A.

Training locally on a small dataset ensures the training script and model parameters are working correctly. Amazon SageMaker training jobs allow direct access to S3 data without downloading everything. Pipe input mode efficiently streams data from S3 to the training instance, reducing disk space requirements and speeding up training.

Only Pipe mode can stream data from S3

The reason for this choice is that Pipe input mode is a feature of Amazon SageMaker that allows you to stream data directly from an Amazon S3 bucket to your training instances without downloading it first1. This way, you can avoid the time and space limitations of loading a large dataset onto your notebook instance. Pipe input mode also offers faster start times and better throughput than File input mode, which downloads the entire dataset before training1.

A. YES - pipe mode is best to start inference before the entire data is transferred; the only drawback is if multiple training jobs are done in sequence (eg. different hyperparamater), the data will be downloaded again B. NO - we want to use SageMaker first for initial training C. NO - We first want to test things in SageMaker D. NO - the SageMaker notebook will not use the AMI so the testing done is useless

B. Generate daily precision-recall data in Amazon QuickSight, and publish the results in a dashboard shared with the Business team. This solution leverages QuickSight's managed service capabilities for both data processing and visualization, which should minimize the coding effort required to provide the Business team with the necessary insights. However, it's important to note that QuickSight's ability to calculate the precision-recall data depends on its support for the necessary statistical functions or the availability of such calculations in the dataset. If QuickSight cannot perform these calculations directly, option C might be necessary, despite the increased effort.

I think it should be a

It's A, pipe mode is for dealing with very big data.

A, PIPE is to do that sort of modeling

When data is already in S3 and next it should move to Sagemaker.. so option A is suitable

Answer is A. B, C & D can be dropped because there is no integration from/to Sage Maker train job (model).

Gotta be A. You need to use Pipe mode but Glue cannot train a model.

AAAAAAAAAAa

ans is A

Will you run AWS Deep Learning AMI for all cases where the data is very large in S3? Also what role is Glue playing here? Is there a transformation? These are the two issues for options B C and D. I believe they do not represent what is required to satisfy the requirements in the question. The answer definitely requires the pipe mode, but not with Glue. I go with A https://aws.amazon.com/blogs/machine-learning/using-pipe-input-mode-for-amazon-sagemaker-algorithms/

go for A