AttributeError: 'MAPPOTrainer' object has no attribute '_local_ip' about marllib

[gyccccccccc]

It's the "D:\py_learning\MARLlib-master\marllib\envs\base_env\exp_results\mappo_mlp_multimerge\MAPPOTrainer_my_merge_multimerge_00c80_00000_0_2023-12-13_18-40-16\params.json"

{
"batch_mode": "truncate_episodes",
"clip_param": 0.3,
"entropy_coeff": 0.01,
"env": "my_merge_multimerge",
"evaluation_interval": 50,
"framework": "torch",
"kl_coeff": 0.2,
"lambda": 1.0,
"lr": 0.0005,
"model": {
"custom_model": "Centralized_Critic_Model",
"custom_model_config": {
"agent_level_batch_update": false,
"agent_name_ls": [
"vehicle1",
"vehicle2"
],
"algorithm": "mappo",
"checkpoint_end": true,
"checkpoint_freq": 50,
"env": "my_merge",
"env_args": {
"map_name": "multimerge"
},
"episode_limit": 100,
"evaluation_interval": 50,
"force_coop": false,
"framework": "torch",
"global_state_flag": true,
"local_dir": "",
"local_mode": true,
"mask_flag": false,
"model_arch_args": {
"core_arch": "mlp",
"encode_layer": "128-128",
"fc_layer": 2,
"hidden_state_size": 256,
"out_dim_fc_0": 128,
"out_dim_fc_1": 64
},
"num_agents": 2,
"num_cpus_per_worker": 1,
"num_gpus": 1,
"num_gpus_per_worker": 0,
"num_workers": 2,
"opp_action_in_cc": true,
"policy_mapping_info": {
"multimerge": {
"all_agents_one_policy": true,
"description": "multi agents merge",
"one_agent_one_policy": true,
"team_prefix": "vehicle_"
}
},
"restore_path": {
"model_path": "",
"params_path": ""
},
"seed": 321,
"share_policy": "all",
"space_act": "Discrete(5)",
"space_obs": "Dict(obs:Box([-inf -inf -inf -inf -inf -inf -inf -inf -inf -inf -inf -inf -inf -inf\n -inf -inf -inf -inf -inf -inf -inf -inf -inf -inf -inf], [inf inf inf inf inf inf inf inf inf inf inf inf inf inf inf inf inf inf\n inf inf inf inf inf inf inf], (25,), float32))",
"stop_iters": 9999999,
"stop_reward": 999999,
"stop_timesteps": 2000000
}
},
"multiagent": {
"policies": "{'shared_policy'}",
"policy_mapping_fn": "<function run_cc.. at 0x0000023968971700>"
},
"num_gpus": 1,
"num_gpus_per_worker": 0,
"num_sgd_iter": 5,
"num_workers": 2,
"seed": 321,
"sgd_minibatch_size": 1000,
"simple_optimizer": false,
"train_batch_size": 1000,
"use_gae": true,
"vf_clip_param": 10.0,
"vf_loss_coeff": 1.0
}

[Aequatio-Space]

The actual error is in the front:

raise ValueError(
ValueError: action_space not provided in PolicySpec for shared_policy and env does not have an action space OR no spaces received from other workers' env(s) OR no action_space specified in config.

It seems that you didn't specify action_space in your environment. You can set relevant properties in custom environment wrapper like:

MARLlib/examples/add_new_env.py

Lines 66 to 79 in 368c617

	class RLlibMAGym(MultiAgentEnv):
	def __init__(self, env_config):
	map = env_config["map_name"]
	env_config.pop("map_name", None)
	self.env = REGISTRY[map](**env_config)
	# assume all agent same action/obs space
	self.action_space = self.env.action_space[0]
	self.observation_space = GymDict({"obs": Box(
	low=0.0,
	high=1.0,
	shape=(self.env.observation_space[0].shape[0],),
	dtype=np.dtype("float64"))})

Or somehow through the configuration file. I set these properties directly through environment code.

[gyccccccccc]

Thank you for your answer :)