Exam AI-100 topic 3 question 10 discussion

will be E. To retrieve logs from a previously deployed web service, experiments are only logging for code to a training script

It is ML Pipelines https://docs.microsoft.com/en-us/azure/machine-learning/concept-ml-pipelines

c. Experiments - Based on the retrieved documents, the Azure Machine Learning service application that you should use to configure versioning and logging for Azure Machine Learning models is Experiments. Experiments allow you to track and manage the versions of your models and log the results of your experiments

pipelines

Deployments is the correct answer

if you download the official Microsoft DP-100 reviewer from GITHUB which you can see here https://github.com/MicrosoftLearning/mslearn-dp100 you can navigate to these notebooks You can register a trained model using 2 approaches 1. using the run.register_model() method which you get when you call experiment.submit() which was used in this notebook 05 - Train Models.ipynb 2. the other method using the Model.register() method which was used in this notebook 08 - Create a Pipeline.ipynb Now regarding logging, you the run object has a number of logging methods 1. run.log() 2. run.log_image() 3. run.log_table() 4. run.log_row() 5. run.log_list() this is all from the "run" object that you get when invoking the experiment.submit() method, and we usually use the run.get_metrics() method to get all the logged metrics using the run.log() method when logging accuracy, f1, etc. on the other hand, MODEL doesn't have any method that "logs" model performance so the answer is definitely EXPERIMENT

The answer is, C. Experiments, https://docs.microsoft.com/en-us/azure/machine-learning/studio/version-control

https://docs.microsoft.com/en-us/azure/machine-learning/how-to-enable-app-insights Deployments??

It should be Models. 'Model registration allows you to store and version your models in the Azure cloud, in your workspace. The model registry makes it easy to organize and keep track of your trained models. After registration, you can then download or deploy the registered model and receive all the files that were registered.' https://docs.microsoft.com/en-us/azure/machine-learning/concept-model-management-and-deployment

for versioning I suppose you have to register the model by using the Model class like shown in the code snippet here https://docs.microsoft.com/en-us/python/api/azureml-core/azureml.core.model.model?view=azure-ml-py and for logging you use Run class

The question is very tricky, as it is asking for logging and versioning for models, not experiment runs. Thinking about model versioning, it reminds me to model registry, which keeps multiple models versions and it comes with Model service: https://docs.microsoft.com/it-it/azure/machine-learning/concept-azure-machine-learning-architecture#models Talking about logging, it could be referred to logs from deployed models, as stated in the answer. Are we sure it is to select only one option? I would have said both Models and Deployments.

I think this would be experiments? https://docs.microsoft.com/en-gb/azure/machine-learning/how-to-enable-logging#logging-for-deployed-models