AI-900 Practice Question: Describe fundamental principles of machine learning on Azure

Answer analysis

Option-by-option breakdown

For each option: why learners choose it and why it is or isn't the right answer here.

• ✗

  An Azure-specific machine learning programming language

  Why it's wrong here

  ONNX is not a programming language of any kind, and it is not tied to Azure. It is an open standard—originally created by Microsoft and Facebook and now governed by the Linux Foundation—that specifies a vendor-neutral graph format and a standardized set of operators for representing trained machine learning models. Azure Machine Learning itself primarily uses Python/R scripts for training and deployment, and other frameworks like PyTorch or TensorFlow can also export/import ONNX models, so its interoperability is cross-cloud and cross-framework.

• ✓

  An open model interchange format enabling models to move between frameworks and edge deployments

  Why this is correct

  ONNX (Open Neural Network Exchange) is an open interchange format for representing a trained ML model as a computational graph with standardized operators, learned parameters, and metadata. This allows a model developed in one framework (e.g., PyTorch) to be converted and run in another runtime (e.g., ONNX Runtime, TensorRT, or Windows ML) with minimal changes, and to be deployed consistently from cloud clusters down to resource-constrained edge devices. Because the format itself is framework- and cloud-agnostic, teams can train once and make the model portable across numerous inference engines, avoiding lock-in to a single vendor. That matches the description in the question exactly.

• ✗

  A database for storing machine learning model training data

  Why it's wrong here

  ONNX is not a data storage service; it is a serialized model format that encodes a computation graph, operator types, and learned weights. Training data typically lives in platform storage like Azure Blob Storage or Azure Data Lake Gen2, or in a database such as Azure SQL or Cosmos DB. While ML pipelines read from those stores to train a model, ONNX simply captures the resulting model for later inference, so confusing it with a training-data database is a category error.

• ✗

  A Microsoft cloud service for distributed model training

  Why it's wrong here

  ONNX is not a Microsoft-managed cloud service, and it performs no distributed training functions. Azure Machine Learning provides the actual infrastructure for large-scale training—such as compute clusters, GPU nodes, and orchestration—whereas ONNX is a file standard that represents an already-trained model. You can save a model from a distributed training job as ONNX and later serve it anywhere, but ONNX itself never allocates compute, schedules tasks, or runs anything; it is simply a portable model file format.

Go deeper

Related to this question