a7582d472ecd556e5f75e18fb3cf44fa5fce3897,ml/rl/training/sac_trainer.py,SACTrainer,train,#SACTrainer#Any#Any#,107

Before Change

        if evaluator is not None:
            cpe_stats = BatchStatsForCPE(
                td_loss=q1_loss.detach().cpu().numpy(),
                logged_rewards=reward.detach().cpu().numpy(),
                model_values_on_logged_actions=q1_value.detach().cpu().numpy(),
                model_propensities=actor_output.log_prob.exp().detach().cpu().numpy(),
                model_values=min_q_actor_value.detach().cpu().numpy(),

After Change

            )
            SummaryWriterContext.add_histogram("actor/loss", actor_loss)

        self.loss_reporter.report(td_loss=float(q1_loss), reward_loss=None)

        if evaluator is not None:
            cpe_stats = BatchStatsForCPE(
                logged_rewards=reward.detach(),

Instances