2a548989f90026395d3d47ccf15ac331728c64bf,ml/rl/test/gym/gym_evaluator.py,GymEvaluator,evaluate_reward_value,#GymEvaluator#Any#,156

Before Change

        if self._env.action_type == EnvType.CONTINUOUS_ACTION:
            raise NotImplementedError()
        // test only float features
        predictions = predictor.predict(self.logged_states).astype(np.float32)
        estimated_reward_values = predictor.estimate_reward(self.logged_states)
        if isinstance(predictor.trainer, ParametricDQNTrainer):
            predictions = predictions.reshape([-1, self._env.action_dim])
            estimated_reward_values = estimated_reward_values.reshape(

After Change

            )
            estimated_reward_values = predictor.estimate_reward(self.logged_states)
        else:
            num_states = self.logged_states.size()[0]
            action_tiled = torch.repeat_interleave(
                torch.eye(self._env.action_dim), repeats=num_states, axis=0
            )

            predictions = predictor.predict(  // type: ignore
                self.logged_states, action_tiled
            )
            estimated_reward_values = predictor.estimate_reward(
                self.logged_states, action_tiled
            )
            predictions = predictions.reshape([-1, self._env.action_dim])
            estimated_reward_values = estimated_reward_values.reshape(
                [-1, self._env.action_dim]
            )

Instances