matesanz / mlops-course

This repository teaches how to train, evaluate and deploy ML models using MLFlow

Hi there 👋

I'm a Computer Vision Developer who tries to make computers see. And I really enjoy it! 😄 I give talks about Artificial Intelligence and I'm really interested in robotics 🦾 I also like board games, doing sports, playing guitar or just being with my friends. ✌️

• 👷 I'm currently working using Computer Vision (classic and deep learning applied)
• 🔭 I’m worked with ROS and applied Artificial Intelligence
• 🐍 I love Python Programming, but I also know C++ and Javascript!
• 📈 I've worked with Kafka and Spark aslo!
• 🌐 Once ago I used to develop VR experiences with Unreal Engine
• 🌱 For my personal projects I've used Streamlit and Vue as frontend and FastAPI/Flask for the backend
• 👨‍🏫 I love teaching and I do it selflessly at Malaga Artificial Intelligence Association
• 🚀 I've also built my own FPV Racing Drone! coool!

mlops-course's Issues

Fix dev container

• In devcontainer.json fix paths to linters.
• Add D401 and D411 Errors to pylint.
• Remove Failfast True
• Add semantic versioning

Log artifacts and models

• Log secondary files from a MLFlow run
• Log main models to MLFlow

Update content for edition II

Objectives

• Upgrade Poetry to 1.6.1
• Move dependencies installation to external post-install.sh
• Remove FastaAPI module
• Upgrade to MLFLOW 2.7.1 (latest currently)
• Remove Pycodestyle dependency due to Poetry >= 1.5 incompatibility
• Document everything

Upgrade Course

• Add Docker section
  • Explain how to install docker
  • Explain what is docker
  • Add hello world docker tutorial
  • Explain how to build your own dockerfile
• Explain how to install mlflow using pip
• Explain how to launch a mlflow server in command line
• Add new sections to MLFLOW:
  • Upgrade to MLFLOW 2.10
  • Explain model signatures

Evaluating a model

Create the API

• Create the API
• Load model
• Perform inference

Add Model Deployment

• Show how to deploy a model using MLFlow
• Explain what an API is
• Explain how to interact with an API

Improve Docs

• Add notebooks to mkdocs
• Change color to teal
• Add CI pipeline to automatically deploy docs on github pages