| import wandb |
|
|
| from src import dataloaders, model, train, utils |
| import pytorch_lightning as pl |
| import logging |
|
|
| def run_pipeline(): |
| config = utils.load_config("gbl_config.yaml") |
| model_config = utils.load_model_config() |
| utils.setup_logging(config["log_level"]) |
| logger = logging.getLogger(__name__) |
| |
|
|
|
|
| pl.seed_everything(model_config["seed"], workers=True) |
|
|
| logger.info("--- Iniciando pipeline de entrenamiento ---") |
|
|
| logger.debug("iniciando Weights & Biases...") |
| wandb.init(project="mdl-mlops", name=f"{model_config['model_name']}-v{model_config['model_version']}", config=model_config, job_type="training") |
| logger.debug("Weights & Biases iniciado") |
|
|
| logger.debug("Definiendo dataloaders y modelo...") |
| dataloader = dataloaders.define_dataloaders(batch_size=model_config["data_batch_size"]) |
| conv_model = model.ConvCVAE(latent_dim=model_config["latent_dim"], lr=model_config["learning_rate"]) |
| logger.debug("Dataloaders y modelo definidos correctamente") |
| |
| logger.debug("Iniciando entrenamiento...") |
| train.train_model(conv_model, dataloader, batch_size=model_config["train_batch_size"], max_epochs=model_config["epochs"], model_name=model_config["model_name"], version=model_config["model_version"]) |
| logger.debug("Entrenamiento finalizado correctamente") |
|
|
| logger.info("--- Pipeline de entrenamiento finalizado ---") |
|
|
| run_pipeline() |
