Training Template

Below is a template in which you should place your model training code. All the outputs from this train script are saved in params["output_dir"].

• Trained model: trained with train data and validated with val data. The model file name and file format are specified, respectively by params["model_file_name"] and params["model_file_format"].

• Predictions on val data: raw model predictions using the trained model on val data. The predictions are saved in val_y_data_predicted.csv.

• Prediction performance scores on val data: calculated using the raw model predictions and the true values for performance metrics specified by metric_type. The scores are saved as json in val_scores.json.

The file should be named mymodel_train_improve.py where ‘mymodel’ is the name of your model.

import sys
from pathlib import Path
from typing import Dict
# [Req] IMPROVE imports
from improvelib.applications.drug_response_prediction.config import DRPTrainConfig
from improvelib.utils import str2bool
import improvelib.utils as frm
from improvelib.metrics import compute_metrics
from model_params_def import train_params
# [MODEL] Model-specific imports, as needed

# [Req]
filepath = Path(__file__).resolve().parent

# [Req]
def run(params: Dict):
    """ Run model training.

    Args:
        params (dict): dict of IMPROVE parameters and parsed values.

    Returns:
        dict: performance scores computed on val data.
    """
    # --------------------------------------------------------------------
    # [Req] Create data names for train/val sets and build model path
    # --------------------------------------------------------------------
    train_data_fname = frm.build_ml_data_file_name(data_format=params["data_format"], stage="train")
    val_data_fname = frm.build_ml_data_file_name(data_format=params["data_format"], stage="val")

    modelpath = frm.build_model_path(
        model_file_name=params["model_file_name"],
        model_file_format=params["model_file_format"],
        model_dir=params["output_dir"])

    # --------------------------------------------------------------------
    # [MODEL] Load model input data (ML data) for train and val
    # --------------------------------------------------------------------

    # --------------------------------------------------------------------
    # [MODEL] CUDA/CPU device, as needed
    # --------------------------------------------------------------------

    # --------------------------------------------------------------------
    # [MODEL] Prepare model
    # --------------------------------------------------------------------

    # --------------------------------------------------------------------
    # [MODEL] Train. Iterate over epochs.
    # --------------------------------------------------------------------

    # --------------------------------------------------------------------
    # [MODEL] Load best model and compute predictions
    # --------------------------------------------------------------------

    # --------------------------------------------------------------------
    # [Req] Save raw predictions in dataframe
    # --------------------------------------------------------------------
    frm.store_predictions_df(
        y_true=your_val_true,
        y_pred=your_val_predicted,
        stage="val",
        y_col_name=params["y_col_name"],
        output_dir=params["output_dir"]
    )

    # --------------------------------------------------------------------
    # [Req] Compute performance scores
    # --------------------------------------------------------------------
    val_scores = frm.compute_performance_scores(
        y_true=your_val_true,
        y_pred=your_val_predicted,
        stage="val",
        metric_type=params["metric_type"],
        output_dir=params["output_dir"]
    )

    return val_scores


# [Req]
def main(args):
    cfg = DRPTrainConfig()
    params = cfg.initialize_parameters(
        pathToModelDir=filepath,
        default_config="MODEL_params.txt",
        additional_definitions=train_params)
    val_scores = run(params)
    print("\nFinished training model.")


# [Req]
if __name__ == "__main__":
    main(sys.argv[1:])