Description:

This course covers in-depth concepts and practical implementation of MLFlow for tracking experiments, managing model registries, tuning hyperparameters, designing MLOps lifecycle, logging model activities, and evaluating model performance. Gain hands-on experience in utilizing MLFlow for efficient machine learning model development and deployment.

Key Highlights:

• Comprehensive MLFlow training
• Hands-on experimentation and model tracking
• Hyperparameter tuning techniques
• Implementing MLOps strategies
• Effective model logging and evaluation

What you will learn:

• Deep dive into MLFlow
  Learn the fundamentals and advanced features of MLFlow for efficient machine learning workflow management.
• Practical experiment tracking
  Hands-on experience in tracking machine learning experiments for better project management and reproducibility.
• Hyperparameter optimization
  Explore various hyperparameter tuning techniques to enhance model performance and efficiency.
• MLOps implementation
  Learn how to incorporate MLFlow into the MLOps lifecycle for streamlined development and deployment processes.
• Model logging and evaluation
  Understand the importance of comprehensive model logging and effective model evaluation techniques.

Curriculum

Session 1 - Introduction to MLFlow and Experiment Tracking

1. What is experimentation tracking, where does it fit in MLOps lifecycle and it's benefits.
2. Tracking through MLFlow.
3. Unique features of MLflow.
4. Why is it used widely in industry?
5. Recent updates

Session -2 Implementing Experiment Tracking through MLFlow

1. Using MLflow in jupyter environment.
2. Logging key parameters and metrics through MLflow.
3. ML flow server UI - detailed walkthrough
4. Model logging.
5. Model Evaluation through MLflow.

Session 3 - Hyperparameter tuning and model registry through MLFlow

1. logging of Hyperparameter tuning runs through Grid search method.
2. Hyperparameter tuning and logging through Hyperopt/Optuna
3. Registering your best model and its version tracking.
4. Model deployment overview through Docker in MLFlow