3e6f4f55b5c967c1f75426260b31109ba3c982e6,reagent/evaluation/ranking_policy_gradient_evaluator.py,RankingPolicyGradientEvaluator,evaluate,#RankingPolicyGradientEvaluator#Any#,48

Before Change

        //  `notify_observers`.
        self.notify_observers(
            eval_baseline_loss=torch.tensor(eval_baseline_loss).reshape(1),
            eval_advantages=torch.FloatTensor(eval_advantage),
            logged_slate_probs=torch.FloatTensor(logged_slate_log_prob),
            ranked_slate_probs=torch.FloatTensor(ranked_slate_prob),
        )

After Change

            .cpu()
        )

        eval_baseline_loss = torch.tensor([0.0]).reshape(1)
        if self.trainer.baseline_net:
            baseline_net = self.trainer.baseline_net
            // pyre-fixme[16]: `Optional` has no attribute `training`.
            baseline_net_prev_mode = baseline_net.training

Instances