utils/hparam_callback.py

from stable_baselines3.common.callbacks import BaseCallback
from stable_baselines3.common.logger import HParam


class HParamCallback(BaseCallback):
    def __init__(self):
        super().__init__()

    def _on_training_start(self) -> None:
        hparam_dict = {
            "algorithm": self.model.__class__.__name__,
            "learning rate": self.model.learning_rate,
            "steps_per_iteration": self.model.n_steps * self.model.n_envs,
            "batch_size": self.model.batch_size,
            "optim_epochs_per_iteration": self.model.n_epochs,
            "gamma": self.model.gamma,
            "gae_lambda": self.model.gae_lambda,
            "ent_coef": self.model.ent_coef,
            "vf_coef": self.model.vf_coef,
        }
        metric_dict = {
            "eval/mean_reward": 0,
            "train/loss": 0,
        }
        self.logger.record(
            "hparams",
            HParam(hparam_dict, metric_dict),
            exclude=("stdout", "log", "json", "csv"),
        )

    def _on_step(self) -> bool:
        return True