45496de13b7932d8ec860da06f14d113000803be,ml/rl/test/gridworld/test_gridworld_pytorch.py,TestGridworld,test_reward_boost,#TestGridworld#,114

Before Change

        )
        self.assertGreater(evaluator.mc_loss[-1], 0.12)

        for _ in range(2):
            for tdp in tdps:
                tdp.rewards = tdp.rewards.flatten()
                tdp.not_terminals = tdp.not_terminals.flatten()
                trainer.train(tdp, None)

        predictor = trainer.predictor()
        evaluator.evaluate(predictor)
        print(
            "Post-Training eval: ",

After Change

        tdps = environment.preprocess_samples(samples, self.minibatch_size)

        with tempfile.TemporaryDirectory() as tmpdirname:
            tmp_path = os.path.join(tmpdirname, "model")
            predictor.save(tmp_path, "minidb")
            new_predictor = DQNPredictor.load(tmp_path, "minidb", False)
            evaluator.evaluate(new_predictor)
        print(
            "Pre-Training eval: ",
            evaluator.mc_loss[-1],
            evaluator.value_doubly_robust[-1],

Instances