fb5d49ee143bf28d78968f5065f8e2f449b3c55b,ml/rl/test/constant_reward/test_constant_reward.py,TestConstantReward,test_trainer_maxq,#TestConstantReward#,28

Before Change

            logger.info("Training..", epoch)
            for tdp in tdps:
                maxq_trainer.train_numpy(tdp, None)
            logger.info("Training epoch", epoch, "average q values",
                  np.mean(workspace.FetchBlob(maxq_trainer.q_score_output)),
                  "td_loss", workspace.FetchBlob(maxq_trainer.loss_blob))

        // Q value should converge to very close to 100
        avg_q_value_after_training = np.mean(
            workspace.FetchBlob(maxq_trainer.q_score_output))

After Change

            logger.info("Training.. " + str(epoch))
            for tdp in tdps:
                maxq_trainer.train_numpy(tdp, None)
            logger.info(
                " ".join(
                    [
                        "Training epoch",
                        str(epoch),
                        "average q values",
                        str(np.mean(workspace.FetchBlob(maxq_trainer.q_score_output))),
                        "td_loss",
                        str(workspace.FetchBlob(maxq_trainer.loss_blob)),
                    ]
                )
            )

        // Q value should converge to very close to 100
        avg_q_value_after_training = np.mean(
            workspace.FetchBlob(maxq_trainer.q_score_output)

Instances