Function transform_actions

eval/run_predict_odyssey.py:155–200

Transform the action space from minicpm towards Odyssey. Args: action: the origin action space. Returns: str: the same action described in odyssey dataset.

(action:dict)

def transform_actions(action:dict) -> str:
    """
    Transform the action space from minicpm towards Odyssey.

    Args:
        action: the origin action space.

    Returns:
        str: the same action described in odyssey dataset.
    """
    action_type:int = action["result_action_type"]

    if action_type == 7:
        return 'PRESS_ENTER' # Not in Odyssey's action space.
    elif action_type == 1:
        return 'NO_ACTION'   # Not in Odyssey's action space.
    elif action_type == 3:
        return f'TYPE: {action["result_action_text"]}'
    elif action_type == 6:
        return 'PRESS_HOME'
    elif action_type == 5:
        return 'PRESS_BACK'
    elif action_type == 7:
        return 'PRESS_ENTER'
    elif action_type == 10:
        return 'COMPLETE'
    elif action_type == 11:
        return 'IMPOSSIBLE'
    elif action_type == 0:
        # Deal with the long press.
        y, x = map(lambda x: round(x * 1000), json.loads(action["result_touch_yx"]))
        return f'LONG_PRESS ({x}, {y})'

    elif action_type == 4:
        # deal with the click or scroll.
        sy, sx =  map(lambda x: round(x * 1000), json.loads(action["result_touch_yx"]))
        ey, ex =  map(lambda x: round(x * 1000), json.loads(action["result_lift_yx"]))

        if sx == ex and sy == ey:
            return f'CLICK: ({ex}, {ey})'
        else:
            direction = get_direction({"x":sx, "y":sy}, {"x": ex, "y": ey})
            return f'SCROLL: {direction.upper()}'

    else:
        raise NotImplementedError (f"No matching type for type {action_type}")


def make_his_idx(episode_dir:str,