Exam DP-100 topic 3 question 21 discussion

C is correct. You may use a yml file to create the Environment and then register it to be used everywhere. If you put the file in the experiment folder you need to create the env every run, against the "least administrative effort" requirement

C is correct

C is the correct answer

C is correct

Registering environment is correct.

In exam 2023-03-23

Creating a config.yaml file is not recommended in this scenario since the experiment is being run on both local and remote compute. It would require copying the file to every location where the experiment is run, which can be a time-consuming administrative task. Using an Estimator with the default packages is not recommended in this scenario since the required packages are not typically installed in a default conda environment.

C. The best solution for this scenario is to create and register an environment that includes the required packages, and use this environment for all experiment runs. This ensures that the experiment runs successfully on both local and remote compute with the least administrative effort. Specifying an environment in the run configuration is necessary to ensure that the correct packages and dependencies are used. If no environment is specified, the default environment may not include the required packages, leading to errors. Creating a virtual machine (VM) as a compute target is not necessary in this scenario since the experiment is being run on local and remote compute, not just remote compute. Also, creating and managing a VM can be a time-consuming administrative task.

C cannot be correct if we are creating an Environment using it for every run?

The case described is for a local Workstation, so the answer is correct. For real-time inferencing service a .yml file is needed

D is correct answer. It may not possible to have an environment with all necessary packages installed. yaml can ensure this.