Introduction to Advanced Model, Experiment, and Artifact Management in Practicus AI with MLFlow
The MLFlow integration of the Practicus AI platform provides a solution that significantly simplifies and optimizes the management process of your machine learning projects.
This guide explains how to use model creation, management of experiments and other features MLFlow offers.
Access to MLFlow Interface:
Within Practicus AI, to access MLFlow:
- Open Explore tab
- Select the MLFlow service defined for you in the working region
- You can see the models and experiments saved in MLFlow here
Saving a created model in MLFLow
- Click on Explore tab
- Make sure your Worker is already selected upper left
- Click Worker Files to load content
- Expand samples folder and click on ice_cream.csv
- Click Load
- Click on the Model button
- Click Advance
- Choose Log Exp. Service as MLFlow Primary
- Click OK
- When the model is created, a plugin will arrive and set the incoming plugin like this
- Click OK
Models, Experiments and Artifacts
- Open the opened MLFLow service in the browser from the tab above
- Find the session you created and open the session
- Here you can see the prt format file, the json containing the model's metadata and the pickle
- Click Parameters
- Back to Session
- Find the session you created and click on the '+' sign under table
- Select the first model under session here
- Click on Metrics and see the error metrics saved in MLFlow:
- Scroll to the bottom of the page and access the model artifacts
Scroll to the bottom of the page and access the model artifacts
Sending an Experiment from Notebook to MLFlow
- Back to Notebook opened after the model
- Run step by step and create exp in step 3
- Update setup params and run the cell
- Now you can setup experiments
-
Run the rest of the steps
-
Save setup
- Open MLFlow in the browser
- Click on the new Experiment and open it
- See the changes here