Q: 5
You plan to build a team data science environment. Data for training models in machine learning
pipelines will
be over 20 GB in size.
You have the following requirements:
Models must be built using Caffe2 or Chainer frameworks.
Data scientists must be able to use a data science environment to build the machine learning
pipelines and train models on their personal devices in both connected and disconnected network
environments.
Personal devices must support updating machine learning pipelines when connected to a network.
You need to select a data science environment.
Which environment should you use?
Options
Discussion
Option A saw a similar question in practice and Studio misses support for Caffe2/Chainer.
B. training jobs need AmlCompute not AKS if it's only set up for inference. I've seen similar wording on other practice sets, pretty sure this is what they're testing for. Let me know if you disagree!
A . Studio's support for extra frameworks is super limited, that's the trap here, so only A lets you run Caffe2 or Chainer locally and offline. Seen this confusion on some practice sets.
A tbh, Azure ML Service is the only one here that supports both Caffe2 and Chainer on personal devices and allows for disconnected work. Studio falls short on framework support, and the others are more for scaling or distributed compute. Someone might argue for Databricks but it doesn't handle offline scenarios well. If anyone disagrees, let me know.
A , Studio just doesn’t natively support Caffe2 or Chainer and gets pretty limited with custom frameworks. Azure ML Service lets you use those frameworks locally or offline and sync back when connected. Anyone disagree?
B tbh, since Machine Learning Studio lets users build pipelines and work offline on personal devices. Studio does have options for custom modules so Caffe2 or Chainer could be installed manually. Not 100% sure if there’s a hidden limitation but the question doesn’t specify. Agree?
B, AKS is set as inference here so can't use for training directly.
Likely A-you can't run Caffe2 or Chainer in Studio easily, that's where B trips people up. I think a lot of folks overlook the custom framework limitation for Studio. If I'm missing something with Databricks or AKS, let me know.
Why are so many picking B? Studio can’t handle Caffe2 or Chainer natively and stumbles with custom frameworks.
Its B for me, since Studio allows building pipelines and seems user-friendly for personal devices, plus you can work offline. Pretty sure you could install extra frameworks if you need Caffe2 or Chainer. Not totally sure if I'm missing an Azure-specific limitation though. Anyone picking something else?
Be respectful. No spam.
