[Feature Request] Make GymLikeEnv compatible with IsaacLab Envs

[yu-fz]

Motivation

The current implementation of GymLikeEnv._step(...) is incompatible with IsaacLab environments due to the conversion to numpy when IsaacLab expects a torch tensor:

    def _step(self, tensordict: TensorDictBase) -> TensorDictBase:
        action = tensordict.get(self.action_key)
        action_np = self.read_action(action)
        reward = 0
        for _ in range(self.wrapper_frame_skip):
            (
                obs,
                _reward,
                terminated,
                truncated,
                done,
                info_dict,
            ) = self._output_transform(self._env.step(action_np))

Solution

Perhaps we can add a config parameter to GymLikeEnv to specify if the underlying environment uses either np.ndarray or torch.tensor types.

Checklist

• I have checked that there is no similar issue in the repo (required)

[vmoens]

Sure I'll work on that!