Spaces:

mnoorchenar
/

AutoMLOps

Sleeping

App Files Files Community

AutoMLOps / README.md

mnoorchenar

Sync from GitHub 56c7c35: README: reliable Star/Fork buttons

67e589d 14 days ago

preview code

Raw History Blame Contribute Delete

15.1 kB

metadata

title: AutoMLOps
colorFrom: purple
colorTo: blue
sdk: docker
app_port: 7860
pinned: false

🤖 AutoMLOps

🤖 AutoMLOps — A full-stack ML experiment tracking and pipeline orchestration platform. Train 50+ algorithms, run visual DAG pipelines powered by Apache Airflow, and manage models — all in one Docker container deployed to HuggingFace Spaces.

Features
Architecture
Getting Started
Docker Deployment
Pages
Pipelines
ML Algorithms
Project Structure
Author
Contributing
Disclaimer
License

✨ Features

🎨 Pipeline Studio	Interactive full-screen DAG canvas with clickable nodes, slide-in config panel, and live execution terminal
✈️ Real Apache Airflow	Pipelines execute as genuine Airflow DAGs with XCom, TaskInstance tracking, and DagRun polling
🤖 AutoML Engine	Automated hyperparameter search across all algorithm categories for classification and regression tasks
📈 MLflow Tracking	Every training run logs parameters, metrics, and model artifacts to a persistent SQLite-backed MLflow store
📦 Model Registry	Register, version, and transition models through Staging → Production → Archived lifecycle stages
🌙 Theme Toggle	Dark and light mode with instant CSS variable switching and localStorage persistence
🐳 Single-Container Deployment	Flask + Airflow Scheduler + SQLite in one Docker image — no external services required

🏗️ Architecture

┌─────────────────────────────────────────────────────────────────────┐
│                           AutoMLOps                                 │
│                                                                     │
│  ┌─────────────┐    ┌──────────────────┐    ┌───────────────────┐  │
│  │  Datasets   │───▶│   MLOps Engine   │───▶│   Flask API       │  │
│  │  (sklearn + │    │  (sklearn /      │    │   Backend         │  │
│  │   custom)   │    │   XGBoost /      │    └────────┬──────────┘  │
│  └─────────────┘    │   LightGBM /     │             │             │
│                     │   MLP)           │    ┌────────▼──────────┐  │
│                     └──────────────────┘    │  Pipeline Studio  │  │
│                                             │  AutoML Page      │  │
│  ┌─────────────┐    ┌──────────────────┐    │  Model Registry   │  │
│  │  MLflow DB  │◀───│ Airflow          │    └───────────────────┘  │
│  │  (SQLite)   │    │ Scheduler        │                           │
│  └─────────────┘    │ (DAGs / XCom /   │                           │
│                     │  TaskInstance)   │                           │
│                     └──────────────────┘                           │
└─────────────────────────────────────────────────────────────────────┘

🚀 Getting Started

Prerequisites

Python 3.11+
Docker (for containerised deployment)
Git

Local Installation

# 1. Clone the repository
git clone https://github.com/mnoorchenar/AutoMLOps.git
cd AutoMLOps

# 2. Create a virtual environment
python -m venv venv
source venv/bin/activate  # Windows: venv\Scripts\activate

# 3. Install app dependencies
pip install -r requirements.txt

# 4. Install Apache Airflow with official constraints
AIRFLOW_VERSION=2.10.4
pip install "apache-airflow==${AIRFLOW_VERSION}" \
    --constraint "https://raw.githubusercontent.com/apache/airflow/constraints-${AIRFLOW_VERSION}/constraints-3.11.txt"

# 5. Initialise Airflow metadata DB
export AIRFLOW_HOME=$(pwd)/airflow_home
export AIRFLOW__CORE__DAGS_FOLDER=$(pwd)/dags
airflow db migrate

# 6. Start the Airflow scheduler (background)
airflow scheduler &

# 7. Run the Flask application
python app.py

Open your browser at http://localhost:7860 🎉

🐳 Docker Deployment

# Build and run
docker build -t automlops .
docker run -p 7860:7860 automlops

# Or deploy directly to HuggingFace Spaces
# Push to a Space with sdk: docker in README frontmatter

The Dockerfile builds a single image containing:

Flask application (served via Gunicorn on port 7860)
Apache Airflow Scheduler (started by start.sh)
All Python dependencies with Airflow constraint-file pinning
Pre-initialised Airflow SQLite metadata DB (airflow db migrate)

📄 Pages

Page	Route	Description	Status
🎨 Pipeline Studio	`/`	Interactive DAG canvas — click nodes to configure and execute pipelines	✅ Live
🤖 AutoML	`/automl`	Automated algorithm search across 50+ models for any dataset	✅ Live
📦 Model Registry	`/models`	Browse registered models, versions, and lifecycle stages	✅ Live

🔗 Pipelines

Three production-quality pipelines are pre-built and immediately executable from the Pipeline Studio:

Training Pipeline (training_pipeline)

Load Data → Validate → Preprocess → Feature Engineering → Train Model → Evaluate → Report → Register → Deploy to Staging

Retraining Pipeline (retraining_pipeline)

Drift Detection → Fetch New Data → Merge Datasets → Retrain Champion → A/B Test → Promote to Production

Data Processing Pipeline (data_pipeline)

Ingest Raw Data → Clean Data → Encode Features → Scale Features → Save to Feature Store

Each pipeline node is clickable in the UI — configurable nodes (dataset picker, algorithm picker) show a purple indicator dot. Execution logs stream live in the built-in terminal panel.

🧠 ML Algorithms

# AutoMLOps Algorithm Registry — 50+ algorithms across 2 tasks
ALGORITHMS = {
    "classification": {
        "Linear Models":          ["Logistic Regression", "Logistic Regression (L1)", "Ridge Classifier",
                                   "SGD Classifier", "Passive Aggressive", "Linear Discriminant Analysis"],
        "Tree-Based":             ["Decision Tree", "Random Forest", "Extra Trees",
                                   "Quadratic Discriminant Analysis"],
        "Ensemble / Boosting":    ["Gradient Boosting", "AdaBoost", "Bagging Classifier",
                                   "XGBoost", "LightGBM"],
        "Support Vector Machines":["SVC (RBF Kernel)", "SVC (Polynomial)", "SVC (Linear)", "LinearSVC"],
        "Probabilistic":          ["Gaussian Naive Bayes", "Bernoulli Naive Bayes", "Complement Naive Bayes"],
        "Instance-Based (KNN)":   ["KNN (k=3)", "KNN (k=5)", "KNN (k=9)"],
        "Neural Networks":        ["MLP (Small)", "MLP (Medium)", "MLP (Deep)"],
    },
    "regression": {
        "Linear Models":          ["Linear Regression", "Ridge Regression", "Lasso",
                                   "ElasticNet", "Bayesian Ridge", "Huber Regressor"],
        "Tree-Based":             ["Decision Tree Regressor", "Random Forest Regressor",
                                   "Extra Trees Regressor"],
        "Ensemble / Boosting":    ["Gradient Boosting Regressor", "AdaBoost Regressor",
                                   "Bagging Regressor", "XGBoost Regressor", "LightGBM Regressor"],
        "Support Vector Machines":["SVR (RBF)", "SVR (Linear)"],
        "Instance-Based (KNN)":   ["KNN Regressor (k=3)", "KNN Regressor (k=5)"],
        "Neural Networks":        ["MLP Regressor (Small)", "MLP Regressor (Medium)"],
    }
}

Built-in Datasets:

Dataset	Task	Samples	Features
Iris Flowers	Classification	150	4
Wine Quality	Classification	178	13
Breast Cancer	Classification	569	30
Diabetes Progression	Regression	442	10
California Housing	Regression	20,640	8

📁 Project Structure

AutoMLOps/
│
├── 📂 dags/                      # Apache Airflow DAG files (parsed by scheduler)
│   ├── __init__.py
│   ├── data_pipeline.py
│   ├── retraining_pipeline.py
│   └── training_pipeline.py
├── 📂 mlops/
│   ├── __init__.py
│   ├── airflow_runner.py
│   ├── algorithms.py
│   ├── datasets.py
│   └── trainer.py
├── 📂 pipelines/
│   ├── __init__.py
│   ├── dag_engine.py
│   └── pipeline_defs.py
├── 📂 static/                    # CSS, JavaScript & assets
│   ├── css/
│   └── js/
├── 📂 templates/                 # HTML templates
│   ├── automl.html
│   ├── base.html
│   ├── dashboard.html
│   ├── experiments.html
│   ├── models.html
│   ├── pipeline.html
│   └── … 1 more
├── 📄 Dockerfile                 # Single-container image (Flask + Airflow)
├── 📄 app.py                     # Flask application entry point + all API routes
├── 📄 docs-template.html
├── 📄 docs.html
├── 📄 readme-template.md
├── 📄 requirements.txt           # Python dependencies
└── 📄 start.sh                   # Startup: Airflow scheduler → Gunicorn Flask

👨‍💻 Author

Mohammad Noorchenarboo

Data Scientist | AI Researcher | Biostatistician

📍 Ontario, Canada 📧 mohammadnoorchenarboo@gmail.com

──────────────────────────────────────

🤝 Contributing

Contributions are welcome! Please follow these steps:

Fork the repository
Create a feature branch: git checkout -b feature/amazing-feature
Commit your changes: git commit -m 'Add amazing feature'
Push to the branch: git push origin feature/amazing-feature
Open a Pull Request

⚠️ Disclaimer

This project is developed strictly for educational and research purposes and does not constitute professional advice of any kind. All datasets used are either synthetically generated or publicly available — no real user data is stored. This software is provided "as is" without warranty of any kind; use at your own risk.

📜 License

Distributed under the MIT License.

_{The name "AutoMLOps" is used purely for academic and research purposes. Any similarity to existing product names or trademarks is entirely coincidental. This project has no affiliation with any commercial entity.}