optuna sciprt

tmp.py

import logging
import math
from typing import Tuple

import optuna
from optuna_dashboard._app import create_app


host = "127.0.0.1"
port = 8080

optuna.logging.set_verbosity(logging.CRITICAL)


def create_optuna_storage() -> optuna.storages.InMemoryStorage:
    storage = optuna.storages.InMemoryStorage()

    # Single-objective study
    study = optuna.create_study(study_name="single-objective", storage=storage)

    def objective_single(trial: optuna.Trial) -> float:
        x1 = trial.suggest_float("x1", 0, 10)
        x2 = trial.suggest_float("x2", 0, 10)
        x3 = trial.suggest_categorical("x3", ["foo", "bar"])
        return (x1 - 2) ** 2 + (x2 - 5) ** 2

    study.optimize(objective_single, n_trials=100)

    # Single objective study with 'inf', '-inf', or 'nan' value
    study = optuna.create_study(study_name="single-inf", storage=storage)

    def objective_single_inf(trial: optuna.Trial) -> float:
        x = trial.suggest_float("x", -10, 10)
        if trial.number % 3 == 0:
            return float("inf")
        elif trial.number % 3 == 1:
            return float("-inf")
        else:
            return x**2

    study.optimize(objective_single_inf, n_trials=50)

    # Single objective pruned after reported 'inf', '-inf', or 'nan'
    study = optuna.create_study(study_name="single-inf-report", storage=storage)

    def objective_single_inf_report(trial: optuna.Trial) -> float:
        x = trial.suggest_float("x", -10, 10)
        if trial.number % 3 == 0:
            trial.report(float("inf"), 1)
        elif trial.number % 3 == 1:
            trial.report(float("-inf"), 1)
        else:
            trial.report(float("nan"), 1)

        if x > 0:
            raise optuna.TrialPruned()
        else:
            return x**2

    study.optimize(objective_single_inf_report, n_trials=50)

    ## Single objective with reported nan value
    # study = optuna.create_study(study_name="single-nan-report", storage=storage)

    # def objective_single_nan_report(trial: optuna.Trial) -> float:
    #    x1 = trial.suggest_float("x1", 0, 10)
    #    x2 = trial.suggest_float("x2", 0, 10)
    #    trial.report(0.5, step=0)
    #    trial.report(math.nan, step=1)
    #    return (x1 - 2) ** 2 + (x2 - 5) ** 2

    # study.optimize(objective_single_nan_report, n_trials=100)

    # Single-objective study with dynamic search space
    study = optuna.create_study(
        study_name="single-objective-dynamic", storage=storage, direction="maximize"
    )

    def objective_single_dynamic(trial: optuna.Trial) -> float:
        category = trial.suggest_categorical("category", ["foo", "bar"])
        if category == "foo":
            return (trial.suggest_float("x1", 0, 10) - 2) ** 2
        else:
            return -((trial.suggest_float("x2", -10, 0) + 5) ** 2)

    study.optimize(objective_single_dynamic, n_trials=50)

    # Single-objective study with 1 parameter
    study = optuna.create_study(
        study_name="single-objective-1-param", storage=storage, direction="maximize"
    )

    def objective_single_with_1param(trial: optuna.Trial) -> float:
        x1 = trial.suggest_float("x1", 0, 10)
        return -((x1 - 2) ** 2)

    study.optimize(objective_single_with_1param, n_trials=50)

    # Single-objective study with 1 parameter
    study = optuna.create_study(study_name="long-parameter-names", storage=storage)

    def objective_long_parameter_names(trial: optuna.Trial) -> float:
        x1 = trial.suggest_float(
            "x1_long_parameter_names_long_long_long_long_long_long_long_long_long_long", 0, 10
        )
        x2 = trial.suggest_float(
            "x2_long_parameter_names_long_long_long_long_long_long_long_long_long_long", 0, 10
        )
        return (x1 - 2) ** 2 + (x2 - 5) ** 2

    study.optimize(objective_long_parameter_names, n_trials=50)

    # Multi-objective study
    study = optuna.create_study(
        study_name="multi-objective",
        storage=storage,
        directions=["minimize", "minimize"],
    )

    def objective_multi(trial: optuna.Trial) -> Tuple[float, float]:
        x = trial.suggest_float("x", 0, 5)
        y = trial.suggest_float("y", 0, 3)
        v0 = 4 * x**2 + 4 * y**2
        v1 = (x - 5) ** 2 + (y - 5) ** 2
        return v0, v1

    study.optimize(objective_multi, n_trials=50)

    # Multi-objective study with dynamic search space
    study = optuna.create_study(
        study_name="multi-dynamic", storage=storage, directions=["minimize", "minimize"]
    )

    def objective_multi_dynamic(trial: optuna.Trial) -> Tuple[float, float]:
        category = trial.suggest_categorical("category", ["foo", "bar"])
        if category == "foo":
            x = trial.suggest_float("x1", 0, 5)
            y = trial.suggest_float("y1", 0, 3)
            v0 = 4 * x**2 + 4 * y**2
            v1 = (x - 5) ** 2 + (y - 5) ** 2
            return v0, v1
        else:
            x = trial.suggest_float("x2", 0, 5)
            y = trial.suggest_float("y2", 0, 3)
            v0 = 2 * x**2 + 2 * y**2
            v1 = (x - 2) ** 2 + (y - 3) ** 2
            return v0, v1

    study.optimize(objective_multi_dynamic, n_trials=50)

    # Pruning with no intermediate values
    study = optuna.create_study(study_name="binh-korn-function-with-constraints", storage=storage)

    def objective_prune_with_no_trials(trial: optuna.Trial) -> float:
        x = trial.suggest_float("x", -15, 30)
        y = trial.suggest_float("y", -15, 30)
        v = x**2 + y**2
        if v > 100:
            raise optuna.TrialPruned()
        return v

    study.optimize(objective_prune_with_no_trials, n_trials=100)

    # With failed trials
    study = optuna.create_study(study_name="failed trials", storage=storage)

    def objective_sometimes_got_failed(trial: optuna.Trial) -> float:
        x = trial.suggest_float("x", -15, 30)
        y = trial.suggest_float("y", -15, 30)
        v = x**2 + y**2
        if v > 100:
            raise ValueError("unexpected error")
        return v

    study.optimize(objective_sometimes_got_failed, n_trials=100, catch=(Exception,))

    # No trials single-objective study
    optuna.create_study(study_name="no trials single-objective study", storage=storage)

    # study with waiting trials
    study = optuna.create_study(study_name="waiting-trials", storage=storage)
    study.enqueue_trial({"x": 0, "y": 10})
    study.enqueue_trial({"x": 10, "y": 20})

    # No trials multi-objective study
    optuna.create_study(
        study_name="no trials multi-objective study",
        storage=storage,
        directions=["minimize", "maximize"],
    )

    # Single-objective study with attrs
    study = optuna.create_study(study_name="single-objective-with-attrs", storage=storage)

    def objective_single(trial: optuna.Trial) -> float:
        x1 = trial.suggest_float("x1", 0, 10)
        x2 = trial.suggest_float("x2", 0, 10)
        x3 = trial.suggest_categorical("x3", ["foo", "bar"])
        trial.set_user_attr("x1", x1)
        trial.set_user_attr("x2", x2)
        return (x1 - 2) ** 2 + (x2 - 5) ** 2

    study.optimize(objective_single, n_trials=100)

    # Single objective study with 'inf', '-inf', or 'nan' value with attrs
    study = optuna.create_study(study_name="single-inf-with-attrs", storage=storage)

    def objective_single_inf(trial: optuna.Trial) -> float:
        x = trial.suggest_float("x", -10, 10)
        trial.set_user_attr("x1", x)
        trial.set_user_attr("x", x)
        if trial.number % 3 == 0:
            return float("inf")
        elif trial.number % 3 == 1:
            return float("-inf")
        else:
            return x**2

    study.optimize(objective_single_inf, n_trials=50)

    return storage


def main() -> None:
    storage = create_optuna_storage()
    app = create_app(storage, debug=True)
    app.run(host=host, port=port, reloader=True)


if __name__ == "__main__":
    main()