feat: implement training script with Hydra configuration

# Standard Imports

import random

# Third-party imports

import hydra

from omegaconf import DictConfig, OmegaConf

# Local imports

from regstar_library.launch.experiment.experiment import Experiment, RayExperiment

from regstar_library.logging.logging import configure_logging

OmegaConf.register_new_resolver(

    "common_seed", lambda: random.randint(0, 2**31), use_cache=True, replace=True

)

@hydra.main(version_base=None, config_path="../conf", config_name="default")

def train(cfg: DictConfig) -> None:

    configure_logging(cfg.logging)

    experiment_logger = hydra.utils.instantiate(

        config=cfg.experiment.logger, _recursive_=True, _convert_="all"

    )

    metrics = hydra.utils.instantiate(

        config=cfg.experiment.metrics, _recursive_=True, _convert_="all"

    )

    if cfg.get("use_ray", False):

        experiment = RayExperiment(

            logger=experiment_logger,

            metrics=metrics,

            run=cfg.experiment.run,

            seed=cfg.experiment.seed,

            epochs=cfg.experiment.epochs,

            ray_kwargs=OmegaConf.to_container(cfg.ray_kwargs, resolve=True),

        )

    else:

        experiment = Experiment(

            logger=experiment_logger,

            metrics=metrics,

            run=cfg.experiment.run,

            epochs=cfg.experiment.epochs,

            seed=cfg.experiment.seed,

        )

    experiment.setup()

    experiment.log_config(cfg.experiment)

    experiment.execute()

    experiment.teardown()

if __name__ == "__main__":

    train()