Stack

The stack within the stack file configuration refers to the MLOps stack intended for deployment. This stack consists of several layers, each aligning with different stages of the MLOps project lifecycle. Some key layers include:

  • data_versioning: For dataset version control.
  • experiment_tracker: To monitor project metadata.
  • orchestrator: To manage data processing, model training, and hyperparameter tuning tasks.
  • model_registry: For tracking and versioning models generated by the orchestrator.

Each layer plays a pivotal role in the streamlined execution and management of MLOps processes.

A sample stack block can look as follows:

Minimal kubernetes based stack for lakefs
stack:
  - data_versioning:
      name: lakefs
Advanced kubernetes based stack for lakefs with many layers of configuration
stack:
  - data_versioning:
      name: lakefs
      params:
        remote_tracking: true
        database_type: "postgres"
        tags:
          database_type: "postgres"
          data_versioning: "lakefs"
          remote_tracking: true

mlinfra is designed to integrate a multitude of stacks, aiming to simplify the deployment process to a mere button click.

Info

mlinfra is under active development and some of the stacks might not be available across all providers and/or deployment. Please keep an eye on the issue tracker and roadmap for more details.