InfinStor Logo

MLflow Service

Start using our hosted MLflow service by purchasing from the AWS Marketplace - free trial included

What is MLflow Service?

InfinStor's MLflow Service is a cloud hosted MLflow service

Easiest way to use MLflow

Our service is securely available 24x7 in the cloud. You can access InfinStor MLflow from anywhere - your laptop, your VMs, your Kubernetes containers, etc. without the need to run your own MLflow server, Database, etc.

Cost

Simple - $120 per year per user. No limits of any kind - experiments, runs, models, uptime, etc. The MLflow database and the MLflow artifact storage S3 bucket are created in your own AWS account - you are responsible for those charges

How to use the MLflow Service

Step by step instructions for using the InfinStor MLflow Service

  1. Set up the MLFLOW_TRACKING_URI environment variable

    • Set the environment variables MLFLOW_TRACKING_URI to infinstor://mlflow.infinstor.com/

      export MLFLOW_TRACKING_URI=infinstor://mlflow.infinstor.com
    • Note that this environment variable must be available in all locations where you plan to use mlflow, e.g. Jupyterlab

  2. Create a new experiment

    • Run the command 'mlflow experiments create -n aabbcc', or alternatively, press the plus button in the MLflow web UI to create a new experiment. Note the experiment id printed out by this command

      mlflow experiments create -n aabbcc
  3. Configure the MLFLOW_EXPERIMENT_ID environment variable

    • The environment variable MLFLOW_EXPERIMENT_ID must be set to the experiment id created in step 2

      export MLFLOW_EXPERIMENT_ID=31
  4. Access the MLflow web UI.

    • The MLflow web user interface is available here.

Step 1: Sign up for the AWS Marketplace Free Trial

The AWS Marketplace Free Trial is a great way to evaluate the InfinStor MLflow Service - detailed instructions here

  1. Login to your AWS account using your web browser

    • Search for Marketplace

  2. In the Marketplace search, look for InfinStor

  3. Sign up for the free trial

Step 2: Configure AWS Account

Configure your own AWS Account for use with InfinStor. Details here

  1. We provide a convenient CloudFormation template that sets up your AWS account

  2. Note that this CFT creates a dynamodb table for storing the MLflow database and an S3 bucket for MLflow artifacts

Step 3: Use InfinStor MLflow

Login and use InfinStor MLflow - details