Introduction

Optuna can print many trial messages, which is useful while debugging but noisy in CI logs and notebooks. Many users look for a direct verbosity=0 argument on study.optimize, but logging control is handled through Optuna’s logging module. The practical solution is to set logger level, disable progress output, and keep only failure-relevant signals.

Where the Noise Comes From

Optimization output usually comes from three sources:

• Optuna logger messages for each trial.
• Progress bar output from show_progress_bar=True.
• Your own print statements in objective functions or callbacks.

If you suppress only one source, logs can still look verbose. Quiet runs require all three to be managed together.

Set Optuna Logger to Warning

The baseline setting is warning level.

import optuna

optuna.logging.set_verbosity(optuna.logging.WARNING)

This hides routine info messages while preserving warnings and errors.

Minimal quiet optimization:

1import optuna
2
3optuna.logging.set_verbosity(optuna.logging.WARNING)
4
5def objective(trial):
6    x = trial.suggest_float("x", -10.0, 10.0)
7    y = trial.suggest_float("y", -10.0, 10.0)
8    return (x - 2.5) ** 2 + (y + 1.0) ** 2
9
10study = optuna.create_study(direction="minimize")
11study.optimize(objective, n_trials=30, show_progress_bar=False)
12
13print("best value:", study.best_value)
14print("best params:", study.best_params)

Disable Progress Bar Explicitly

Even with warning-level logs, progress bars can still write frequent output. Disable them for clean batch execution.

study.optimize(objective, n_trials=100, show_progress_bar=False)

This is especially important in CI environments where progress bar redraw produces large logs.

Keep Objective and Callback Functions Quiet

A common mistake is suppressing Optuna logs but leaving debug prints in objective code.

1DEBUG = False
2
3def objective(trial):
4    lr = trial.suggest_float("lr", 1e-4, 1e-1, log=True)
5    if DEBUG:
6        print("trial", trial.number, "lr", lr)
7    return (lr - 0.01) ** 2

Also audit callback functions for per-trial prints.

Environment-Driven Logging Policy

Use environment variables to switch between quiet production mode and verbose debugging mode.

1import os
2import optuna
3
4if os.getenv("OPTUNA_DEBUG") == "1":
5    optuna.logging.set_verbosity(optuna.logging.INFO)
6else:
7    optuna.logging.set_verbosity(optuna.logging.WARNING)

This keeps one code path while allowing flexible runtime behavior.

Integrate with Python Logging

If your app configures Python logging globally, align levels so third-party logs remain predictable.

1import logging
2import optuna
3
4logging.basicConfig(level=logging.WARNING)
5optuna.logging.set_verbosity(optuna.logging.WARNING)

Do this before running optimization jobs, not after they start.

What About True Silent Mode

If you need almost no output:

• Set Optuna to warning.
• Disable progress bar.
• Remove objective prints.
• Print only final metrics.

This gives near-silent behavior without hiding critical failures.

For incident triage, temporarily raise logger to info level and rerun a small trial count.

Callback and Multi-Worker Considerations

In distributed or parallel runs, custom callbacks can reintroduce high-volume output even when Optuna logger is quiet. Keep callback logging rate-limited and write detailed per-trial diagnostics to files instead of standard output.

def callback(study, trial):
    if trial.number % 20 == 0:
        print("checkpoint trial:", trial.number, "best:", study.best_value)

This keeps console output compact while preserving periodic visibility during long studies.

Common Pitfalls

• Expecting study.optimize to have a universal verbosity=0 switch. Fix by configuring Optuna logger directly.
• Forgetting show_progress_bar=False. Fix by disabling progress output in non-interactive runs.
• Leaving debug prints inside objective functions. Fix by guarding prints behind a debug flag.
• Muting all logs and missing trial failures. Fix by keeping warning level instead of silencing everything.
• Applying logger config after optimization starts. Fix by setting logging before study.optimize call.

Summary

• Optuna verbosity control is primarily logger configuration, not a single optimize argument.
• Use optuna.logging.set_verbosity(optuna.logging.WARNING) for quiet default runs.
• Disable progress bars and debug prints for clean logs.
• Use environment-based toggles to switch verbosity safely.
• Preserve warnings so failures remain visible in production workflows.