MLflow

FAQs

• Q. What is MLflow?
  MLflow is an open-source platform for managing machine learning experiments and models. It helps you track, manage, and reproduce the experimental data, code versions, hyperparameters, model results, and deployments that happen when you create ML models in a single system. It makes ML projects more organized and collaborative for developers and data scientists.
• Q. What does MLflow do?
  There are four main components: - Tracking - to record parameters, metrics, result files, etc. for experiments - Projects - a structured way to reproduce code, environments, and dependencies - Models - to store and deploy models of different frameworks - Registry - to manage models as production/test versions and apply approval flows
• Q. Is MLflow free?
  MLflow itself is free. MLflow is an open-source machine learning operations tool that is free for anyone to use under the Apache 2.0 license. However, you will need to configure your own server, storage, DB, etc. and you will be responsible for the infrastructure costs.
• Q. Does it cost money to use MLflow on Databricks?
  MLflow is provided by Databricks, and while you can try it for free for 14 days, you have to pay for full use. Databricks integrates MLflow with a web UI, automatic experiment tracking, model serving, and more. You are billed on a per DBU (Databricks Unit) basis. You only pay for what you use, and there is a 14-day free trial.
• Q. If I use MLflow on Databricks, do I need to set up my own DB or storage?
  No, you don't need to configure any DB or storage in Databricks. Databricks is a serverless environment that automatically manages the storage of experiment information, models, and artifacts. All you need to do is run ML experiments or record them in a notebook - all these infrastructure resources are included and billed in the DBU unit price.
• Q. What is the pricing for MLflow on Databricks?
  Pay-per-use (DBU) billing: - CPU-based: approximately $0.15 to $0.40/DBU - Plans with advanced features: $0.55/DBU and above - GPU, model serving, SQL warehouse, etc. are separately priced - 14-day free trial, after which you must register a payment method to continue using.
• Q. Can MLflow be used without Databricks?
  MLflow is an open source platform that can be installed and used alone under the Apache 2.0 license. It can be freely deployed in local environments, cloud VMs, on-premises servers, and more. Databricks integrates and extends MLflow for enterprise environments.
• Q. Is there a charge for using the MLflow API?
  No. MLflow's own API is free. There is no additional charge to use the REST API or Python API in the open source installations. However, in Databricks environments, computing resource usage for API calls may be reflected in the charges.