search

jidiai / GRF_MARL

Google Research Football MARL Benchmark and Research Toolkit
https://grf-marl.readthedocs.io/
Other
27 stars 4 forks source link

Running Trained models with 'play_with_ai or play_gr_football.py #1

Open synopi opened 8 months ago

synopi commented 8 months ago

I'm having issues playing against the trained model, i've tried the below methods but haven't been successful. Can y'all point me in the right direction please. This is after training using the mappo config

python light_malib/scripts/play_gr_football.py --model_0 "/home/Desktop/GRF_MARL/logs/gr_football/full_game_5_vs_5_hard_mappo/2023-11-12-20-49-10/agent_0/agent_0-default-1/best" [2023-11-13 15:06:39,267][WARNING] use model type: gr_football.basic_centralized_critic_pool_obs [2023-11-13 15:06:39,296][WARNING] use model type: gr_football.built_in_5 Traceback (most recent call last): File "Desktop/GRF_MARL/light_malib/scripts/play_gr_football.py", line 91, in env = GRFootballEnv(0, None, cfg.rollout_manager.worker.envs[0]) File "Desktop/GRF_MARL/light_malib/envs/gr_football/env.py", line 57, in init assert official_reward in ["scoring","checkpoints"],official_reward AssertionError: checkpoints

and,

./play_with_ai.sh AssertionError: Player type 'agent' can not be used with play_game. Execution time was 0 seconds.