launcher.py

import subprocess
import shutil
import json
from pathlib import Path


VERSION_LIST = []


def training(version):
    """
    """
    subprocess.run(["python", "src/launch_train.py", version])

    dataset_dir = Path(__file__).parent / "kernel" / "input" / f"model{version}_aptos2019"

    # mkdir for dataset
    dataset_dir.mkdir(parents=True, exist_ok=True)

    for f_path in list((Path(__file__).parent / "model" / "qwkcoef").glob(f"{version}*.txt")):
        shutil.copy(str(f_path), str(dataset_dir / f_path.name))

    for f_path in list((Path(__file__).parent / "model").glob(f"{version}*.pth")):
        shutil.copy(str(f_path), str(dataset_dir / f_path.name))

    for f_path in list((Path(__file__).parent / "config").glob(f"{version}*.yml")):
        shutil.copy(str(f_path), str(dataset_dir / f_path.name))

    # set metadata
    subprocess.run(["kaggle", "datasets", "init", "-p", f"kernel/input/model{version}_aptos2019"])

    with open(dataset_dir / "dataset-metadata.json") as j:
        metadata = json.load(j)

    metadata["title"] = f"model{version}_aptos2019"
    _id_root = metadata["id"].split("/")[0]
    metadata["id"] = f"{_id_root}/model{version}_aptos2019"

    with open(dataset_dir / "dataset-metadata.json", mode="w") as j:
        json.dump(metadata, j, indent=2)

    # upload model
    subprocess.run(["kaggle", "datasets", "create", "-p", f"kernel/input/model{version}_aptos2019"])


def inference(version):
    """
    """
    notebook_dir = Path(__file__).parent / "kernel" / version
    # notebook_title = f"{version}_aptos2019"

    # mkdir for dataset
    notebook_dir.mkdir(parents=True, exist_ok=True)

    # set metadata
    subprocess.run(["kaggle", "kernels", "init", "-p", f"kernel/{version}"])

    with open(notebook_dir / "kernel-metadata.json") as j:
        metadata = json.load(j)

    username = metadata["id"].split("/")[0]
    metadata["id"] = username + "/" + version + "-aptos2019"
    metadata["title"] = f"{version} aptos2019"
    metadata["code_file"] = "inference.py"
    metadata["language"] = "python"
    metadata["kernel_type"] = "script"
    metadata["is_private"] = "true"
    metadata["enable_gpu"] = "true"
    metadata["enable_internet"] = "false"
    metadata["dataset_sources"] = [username + "/" + f"model{version}_aptos2019", username + "/efficientnet"]
    metadata["competition_sources"] = ["aptos2019-blindness-detection"]
    metadata["kernel_sources"] = []

    with open(notebook_dir / "kernel-metadata.json", mode="w") as j:
        json.dump(metadata, j, indent=2)


def main():
    for v in VERSION_LIST:
        training(v)
        inference(v)


if __name__ == "__main__":
    main()