| import optuna | |
| from optuna.integration import TensorBoardCallback | |
| def save_trial_callback(study, trial, trials_result_path): | |
| with open(trials_result_path, "a") as f: | |
| f.write(f"Trial {trial.number}: Value (F1 Macro): {trial.value}, Params: {trial.params}\n") | |
| def create_optuna_study(objective, n_trials, trials_result_path, tensorboard_log_dir): | |
| study = optuna.create_study(direction="maximize") | |
| # init TensorBoard callback | |
| tensorboard_callback = TensorBoardCallback(dirname=tensorboard_log_dir, metric_name="F1 Macro") | |
| # callback and TensorBoard callback | |
| callbacks = [ | |
| lambda study, trial: save_trial_callback(study, trial, trials_result_path), | |
| tensorboard_callback | |
| ] | |
| study.optimize(objective, n_trials=n_trials, callbacks=callbacks) | |
| return study |