Exam AWS Certified Machine Learning - Specialty topic 1 question 202 discussion

Should be D. TFRecord could be uploaded to S3 directly and be used as SageMaker's data source. https://sagemaker-examples.readthedocs.io/en/latest/sagemaker_batch_transform/working_with_tfrecords/working-with-tfrecords.html#Upload-dataset-to-S3

It has to option D.

A. NO - SageMaker can use TFRecods as-is in S3 B. NO - SageMaker can use TFRecods as-is in S3 C. NO - SageMaker must use S3 as input, it cannot read your local data D. YES - SageMaker can use TFRecods as-is in S3

SageMaker script mode allows you to use your existing TensorFlow training scripts without any modifications. You can use the same TFRecord data format that your model expects, and point the SageMaker training invocation to the S3 bucket where the data is stored. SageMaker will automatically download the data to the local path of the training instance and pass it as an argument to your train.py script. You don’t need to reformat the data to protobuf format or rewrite your script to convert the data12.

This option allows the ML specialist to use the existing train.py script and TFRecord data without any changes, minimizing development overhead. By using SageMaker script mode, the specialist can run the existing TensorFlow script as-is, and by pointing the SageMaker training invocation to the S3 bucket containing the TFRecord data, the specialist can provide the training data to SageMaker without reformatting it.

This option leverages SageMaker's built-in support for the TensorFlow framework and script mode. The existing train.py script can be used without any modifications. SageMaker will automatically download the training data from the specified S3 location to the instance running the training job. This option saves development time by avoiding the need to rewrite the train.py script or reformat the training data.