Tracking machine learning models is a crucial aspect in companies working with unstable data, such as live customer data or any data that changes over time. The ML model might perform well during training in a Jupyter notebook, but it could show poor performance in prediction with real data.

Experiment tracking involves keeping track of all relevant information about an ML model.

ML experiment — This refers to our machine learning model. When someone says they are tracking their experiment, they mean they are training the machine learning model and checking its performance.

Experiment run — Each trial in your machine learning model. For example, you run your machine learning model and get your r2 result. After that, you change some parameters and run again, resulting in 2 runs, each with different parameters and results.

Run artifact — This includes all kinds of information that we want to save with the machine learning model, such as the data source, developer, or environment settings, etc.

Source code, Environment, Data, Model, Hyperparameters, and Metrics are the common choices that most machine learning developers want to track.

The concept of model tracking, the tools, and the methods of how we can track the model are also super important. The tracking should be simple and compact so that not just a senior Machine learning developer can understand the result, but also a junior Machine learning model who started the job a week ago can understand the result. They should be able to get insights from the tracking about which run was successful, which parameters were used, and finally, which machine learning model has the best score. Therefore, it’s time for us to talk about MLflow.

MLflow is an open-source platform for the machine learning lifecycle. It is a python package that you can easily install with the pip command, htt

pip install mlflow

It contains four main modules:

*Tracking
*Models
*Model Registry
*Projects
Tracking experiments with MLflow

The MLflow Tracking module allows you to organize and keep track of:

Parameters — It could be any that influences the model score, such as Hyperparameter but also the data, the source data. It could be, for example, the different version of the data. The model has a good score with a month previous data. However, with the new data, the model shows a bad score.

Metrics — Any metric that helps to score the model performance such as r2, RMSE, etc.

Metadata — Metadata is generally used to add extra information to the model that helps to find the information about this model easily, for example, the tag, the name of the developer could be as a tag.

Artifacts — It helps to interpret the model in an easy way. For example, the graph shows which model is better, or how the performance model is changing over time.

Models — The model that you trained and want to save the model for automation.

Along with this information, MLflow automatically logs extra information about the run:

Source code Version of the code Start and end time Author Let us start with MLflow Create a new repository in the GitHub and go to the code and choose the CodeSpaces.

In the CodeSpace click on three dots and choose Open in Visual Studio Code. When Visual Studio Code starts, you can access your virtual machine locally and develop your model in this virtual machine.

Open the Visual Code terminal (Ctrl+Shift+P) and check the python version:

python — version

Go to the GitHub repositories https://github.com/HuseynA28/mlops_zoombootcamp/tree/main/02-experiment-tracking and install the requirements.txt.

pip install -r requirements.txt

Requirements data contains the latest version of the MLflow and other packages that we will use later.


This content will provide a well-structured `README.md` for your GitHub repository, including code blocks for commands, which improves the document's usability and readability.