CUDA Memory Issue

[NUSShao]

Hi I'm suffering from CUDA out of memory issue, is there a way to intialize the habitat VectorEnv using CPU?

    vector_env_cls: Type[Any]
    if int(os.environ.get("HABITAT_ENV_DEBUG", 0)):
        logger.warn(
            "Using the debug Vector environment interface. Expect slower performance."
        )
        vector_env_cls = ThreadedVectorEnv
    else:
        vector_env_cls = VectorEnv
    # Here using CUDA Memory
    envs = vector_env_cls(
        make_env_fn=make_gym_from_config,
        env_fn_args=tuple((c,) for c in configs),
        workers_ignore_signals=workers_ignore_signals,
    )
    return envs