This contains a repository of all the MLFlow examples in Python. The project structure currently contains:
- Basic ML Flow - checking ML flow version
- Basic Metric Logging
- Logging a Sci-Kit Learn project, alongide model registration with log_model
- Conda packaging - deploy your model in a package and pass to other users, or deploy in the cloud with Azure or GCP
- Specifying additional pip installs - this shows how to build additional requirements into your script
- Working with PyTorch - to create, and package, a MNIST computer vision classification algorithm and s
- Working with Tensorflow 2 model and packaging it up to work with MLFlow
- XGBoost native and XGBoost Scikit Learn - shows different approachesd to packaging these models up
- Dockerising MLFlow - uses MLFlow guides to show how the model can be packages up into Docker and ran as a microservice.
- FastAI Example with MLFlow - how to work with FastAI
- Experiment tracking - shows how to track experiments in MLFlow
- Multiple workstep example in MLFlow - how to work with multiple workflow steps, as you may want to log metrics from data prep, training and evaluation phases.
- Hyperparameter tuning and MLFlow - this shows how to log multiple runs when capturing the hyperparameters.
-
- Model Registration - shows how to register a model using a run ID, experiement ID or in script.
-
- Command Line Interface (CLI) for GCP - aids for working with the command line interface in Google Cloud Platform. To the left there is a menu indicating other cloud deployment options and how to work with MLFlow.
I aim to add a couple more modules on model registration, experiment creation and doing all the steps in script. Watch this space for future additions to this repository.