Exam AWS Certified Machine Learning - Specialty topic 1 question 198 discussion

yes, It seems to be 'B'

B is correct, from within single endpoint, we can create multiple production variant. When lambda called, it should have been each target variant instead of production variant in the verbiage

C is fine B is not possible as it is single sagemaker endpoint (so we won't get prediction from all models for each document) D is wrong as we do not need three lambda functions A is also wrong as time gap is 3 seconds for which we should be running batch transform jobs

Will go with B

A. NO - you don't want to create a new job for each Lambda invokation B. YES - best practice C. NO - could work but does not leverage production variants which in-turn disable some built-in model performance evaluation features D. NO - more operationnal overhead to have 3 endpoints

Answer is B

Although C sounds like a better option but B is less operational overhead at least for short term.

It's B, you can use Invoke a Multi-Model Endpoint, when you call invoke_endpoint you need to provide which model filw to use. response1 = runtime_sagemaker_client.invoke_endpoint( EndpointName = "MAIN_ENDPOINT", TargetModel = "model1.tar.gz", Body = body) response2 = runtime_sagemaker_client.invoke_endpoint( EndpointName = "MAIN_ENDPOINT", TargetModel = "model2.tar.gz", Body = body) response3 = runtime_sagemaker_client.invoke_endpoint( EndpointName = "MAIN_ENDPOINT", TargetModel = "model3.tar.gz", Body = body) Ref: https://docs.aws.amazon.com/sagemaker/latest/dg/invoke-multi-model-endpoint.html

B - the reason is not shadow testing since it is not named and does not require client logic. The reason is that it is possible to target a model https://docs.aws.amazon.com/sagemaker/latest/dg/invoke-multi-model-endpoint.html

B it is which involves using single endpoint for multiple model versions

I think the answer should be C. As there is no production version of the model identified, all the 3 models need to be invoked.

C, prod variant is used for traffic routing. All model needs to be invoked.

C is correct

I think the answer is B.