In what format does MLflow store model artifacts?

MLflow stores model artifacts as files in a cloud storage system or a local filesystem. This choice is correct because MLflow is designed to manage the machine learning lifecycle, including model versioning, and it does so by saving model artifacts—such as model files, data files, and evaluation metrics—in easily accessible formats.

The flexibility of using cloud storage systems like AWS S3, Azure Blob Storage, or local filesystems allows users to store large amounts of data efficiently, supporting different types of models and frameworks. The ability to utilize standard file formats facilitates easy manipulation and retrieval of artifacts, making it easier for teams to collaborate on machine learning projects and deploy models.

In contrast, the other options suggest formats that do not align with how MLflow handles model artifacts. SQL databases, for instance, are not typically utilized for storing model artifacts themselves, as they are more suited for structured data rather than the diverse types of files associated with models. Custom-built file formats and plain text files imply restrictive storage methods that do not take advantage of the flexibility and efficiency offered by cloud storage systems or local filesystems. Therefore, the chosen option exemplifies the practical approach MLflow takes in managing artifacts for machine learning models.