I absolutely need to be able to write parameters in a job. I know it is not the protocol, but is there a way to do it? Thanks! @job def my_test_job(): result_1 = my_op1("value") result_2 = my_op2(20)

I used input configs

train_job = training.to_job(
    resource_defs={
        "bq": bigquery_resource,
        "io_manager": gcs_parquet_pickle_io_manager,
        ...
    },
    config={
        "inputs": {
            "grouping": "brand",
            "time_interval": "daily",
            "split_date": f"{datetime.now() - timedelta(days=14):%Y-%m-%d}",
            "forecast_horizon_days": 30,
            "confidence_interval": 0.95,
            "skip_train": False,
        },
        "resources": {
            "io_manager": IO_CONFIG_GCS,
            ...
        },
    },
)

with this graph

@graph
def training(
    grouping: str,
    time_interval: str,
    split_date: str,
    forecast_horizon_days: int,
    confidence_interval: float,
    skip_train: bool,
):
    """Train a single model and register it"""
    train_df, test_df, train_query, test_query = feature_gen(
        grouping, time_interval, split_date
    )
    model = fit_model(
        train_query, test_query, grouping, time_interval, skip_train
    )
    forecast_df = forecast_unlogged(model, forecast_horizon_days, confidence_interval)
    eval_df = evaluate(model, test_df, forecast_df, grouping)
    success, _ = model_validation(prev=eval_df)
    model_uri = log_model(model, success)
    return model

so the job gets inputs injectes as paramters. alternatively you can have one op which spits out those params