Instruction file imported from abpai/experiments.hydra (.cursor/rules/project-overview.mdc). Copyright stays with the author.

Hydra Experiments Project Overview

This project demonstrates using Hydra for ML experiment configuration management with a spam classification example.

Project Structure

• train.py: Main script for training and evaluating models
• conf/: Contains all Hydra configuration files
  • conf/config.yaml: Main configuration file
  • conf/model/: Model-specific configurations
  • conf/feature_extractor/: Feature extraction configurations
  • conf/dataset/: Dataset configurations

Key Concepts

• Hydra loads configurations from YAML files and composes them
• The main script uses @hydra.main decorator
• Models are dynamically instantiated based on configuration
• Configurations can be overridden via command line

Example Usage

# Run with default configuration
python train.py

# Run with a different model
python train.py model=logistic_regression

# Override specific parameters
python train.py model.alpha=0.5

See Hydra Configuration for more details on configuration.