RL using IPPO on SKRL

I run through this error when i use Petting zoo wrapper to solve a multiagent env and ippo training script

I have my observation space and possible_ agents defined in the environment as:

self.observation_space = spaces.Box(low=float("inf"), high=float("inf"), shape=(16,), dtype=np.float32)
self.possible_agents = ["my_jetbot", "my_jetbot1"]


[skrl:INFO] Environment wrapper: Petting Zoo
Traceback (most recent call last):
  File "torch_isaacsim_jetbot_ippo.py", line 59, in <module>
    env = wrap_env(env, wrapper="pettingzoo")
  File "/home/marty/.local/share/ov/pkg/isaac_sim-2022.2.1/standalone_examples/api/omni.isaac.jetbot/skrl_example/src/skrl/skrl/envs/wrappers/torch/__init__.py", line 126, in wrap_env
    return PettingZooWrapper(env)
  File "/home/marty/.local/share/ov/pkg/isaac_sim-2022.2.1/standalone_examples/api/omni.isaac.jetbot/skrl_example/src/skrl/skrl/envs/wrappers/torch/pettingzoo_envs.py", line 22, in __init__
    self._shared_observation_space = self._compute_shared_observation_space(self._env.observation_spaces)
AttributeError: 'JetBotEnv' object has no attribute 'observation_spaces'

Without having to share my code snippets, could you tell me what i might be doing wrong or missing?

Hi @toni.sm - Can you help the user here?

Hi @qwe.marty

Sorry for late response, was on a family trip.

Please, note that the skrl’s multi-agent environment interface is based on PettingZoo Parallel API.

PettingZoo’s interface differs, in many ways, from Gym/Gymnsium interface.
In PettingZoo Parallel API, the observation_space and actions_space are class methods rather than class properties.

Hey @toni.sm i have posted an updated issue here on Github discussion
Let me know thanks