MLflow Global Setup

Containerized MLflow tracking server with persistent storage.

Quick Start

# Build image
sh scripts/build_image.sh

# Stop and Build image
uv run scripts/stop_container.py && sh scripts/build_image.sh

# Start container (port 5500)
uv run scripts/run_container.py

# Configure shell
source mlflow_env.sh

UI: http://localhost:5500

Usage

Direct logging:

uv run scripts/test_mlflow_project.py

Decorator:

from mlflow_utils import mlflow_track

@mlflow_track(experiment_name="My_Experiment")
def train():
    return {"params": {...}, "metrics": {...}}

Test Decorator

source mlflow_env.sh && uv run scripts/test_decorator_example.py
OR
source mlflow_env.sh && python scripts/test_decorator_example.py

Commands

# Stop container
uv run scripts/stop_container.py
OR
python scripts/stop_container.py

# Check port
lsof -i :5500

Config

• MLflow version: pyproject.toml
• Port/paths: mlflow.env
• Data: ~/mlflow (default)