Introduction to Hydra-Core

Hydra-core is an innovative framework designed to empower your Python projects with flexible and efficient configurations. The library is highly praised for its ability to manage complex parameter configurations, enabling developers to write clean and maintainable code. Here, we dive deep into hydra-core functionalities, APIs, and provide useful examples to help you leverage its full potential.

Getting Started

To get started with Hydra, you need to install it:

  pip install hydra-core --upgrade

Core APIs and Examples

Basic Initialization

You can initialize Hydra in your main function to handle configurations:

  
  from omegaconf import DictConfig
  import hydra
  
  @hydra.main(config_path=None, config_name="config")
  def my_app(cfg: DictConfig):
      print(cfg.pretty())
  
  if __name__ == "__main__":
      my_app()
  

Default Configuration File

Create a configuration file named config.yaml:

  
  database:
    host: localhost
    port: 3306
  

Accessing Configuration

You can access configuration parameters like this:

  
  @hydra.main(config_path=".", config_name="config")
  def my_app(cfg: DictConfig):
      print(f"Connecting to database {cfg.database.host}:{cfg.database.port}")
  
  if __name__ == "__main__":
      my_app()
  

Overriding Configuration

Override configuration from the command line:

  python my_app.py database.port=1234

Composable Configurations

Use multiple configuration files with composition:

  
  # structure of config directory
  # config/
  #   config.yaml
  #   db/mysql.yaml
  
  # config/config.yaml
  defaults:
    - db: mysql
  
  # config/db/mysql.yaml
  database:
    driver: mysql
    host: localhost
    port: 3306
  

Configuration Groups

Manage configuration groups and override them as needed:

  
  @hydra.main(config_path="config", config_name="config")
  def my_app(cfg: DictConfig):
      print(f"Database driver: {cfg.database.driver}")
  
  if __name__ == "__main__":
      my_app()
  

Using Structured Configs

Define structured configuration with classes:

  
  from dataclasses import dataclass
  from omegaconf import MISSING
  import hydra
  
  @dataclass
  class DatabaseConfig:
      driver: str = MISSING
      host: str = MISSING
      port: int = MISSING
  
  @dataclass
  class Config:
      database: DatabaseConfig = DatabaseConfig()
  
  @hydra.main(config_path=None, config_name="config")
  def my_app(cfg: Config):
      print(cfg.database)
  
  if __name__ == "__main__":
      my_app()
  

Integrating with Other Libraries

Hydra can be easily integrated with optuna for hyperparameter optimization:

  
  import optuna
  from hydra import initialize, compose
  
  def objective(trial):
      lr = trial.suggest_float("lr", 1e-4, 1e-1, log=True)
      # Your model training with `lr`
  
  with initialize(config_path="conf"):
      cfg = compose(config_name="config")
      study = optuna.create_study(direction="minimize")
      study.optimize(objective, n_trials=100)
  

Example Application

Here is a more complete example that demonstrates different Hydra capabilities:

  
  # config/config.yaml
  defaults:
    - db: mysql
  
  database:
    driver: mysql
    host: localhost
    port: 3306
  
  server:
    host: 127.0.0.1
    port: 8080
  
  # config/db/mysql.yaml
  database:
    driver: mysql
    host: mysql.local
    port: 3306
  
  # Run Application
  import hydra
  from omegaconf import DictConfig
  from flask import Flask
  
  app = Flask(__name__)
  
  @hydra.main(config_path="config", config_name="config")
  def run(cfg: DictConfig):
      app.run(host=cfg.server.host, port=cfg.server.port)
  
  if __name__ == "__main__":
      run()
  

Hydra-core simplifies the management of configuration complexities in your Python projects. With these insights and examples, you can make your applications more flexible, maintainable, and scalable.

Hash: b66371a4bd5cf200cc3b7f23556c6fcb4f9761f313136e260c5f19a53b634f69