model_name = input('Enter model name: ')

FOLDER = f'models/{model_name}'

if not os.path.exists(FOLDER):

print('Model does not exist')

sys.exit()

params = {}

env_config = {}

# Load params and env_config and scores

with open(f'{FOLDER}/params.json', 'r') as f:

params = json.load(f)

with open(f'{FOLDER}/cf.json', 'r') as f:

env_config = json.load(f)

env_config['show'] = False

env = battle_env.parallel_env(**env_config)

red_agent_list = env.possible_red

blue_agent_list = env.possible_blue

obs_len = env.observation_space(red_agent_list[0]).shape[0]

critic_dims = obs_len * env.n_agents

# Red team is the maddpg team

red_team = maddpg.Team(red_agent_list, obs_len, env.n_actions, critic_dims, params['fc1_dims'], params['fc2_dims'], params['buffer_size'], params['batch_size'], params['gamma'], params['lr'], FOLDER)

red_team.load_models()

# Blue team is the instinct team

blue_team = instinct.Team(blue_agent_list, red_agent_list, env)

wins = {

"red": 0,

"blue": 0,

"tie": 0

}

for _ in range(10000):

observations = env.reset()

red_obs = {}

blue_obs = {}

red_obs_ = {}

blue_obs_ = {}

for agent in red_agent_list:

red_obs[agent] = observations[agent]

for agent in blue_agent_list:

blue_obs[agent] = observations[agent]

observations = env.reset()

actions = {}

while not env.env_done:

actions = merge_dicts(red_team.choose_actions(red_obs), blue_team.choose_actions(blue_obs))

observations_, rewards, dones, _ = env.step(actions)

for agent in red_agent_list:

red_obs_[agent] = observations_[agent]

for agent in blue_agent_list:

blue_obs_[agent] = observations_[agent]

red_obs = red_obs_

blue_obs = blue_obs_

wins[env.winner] += 1

env.show = True

env.start_recording(path=f'{FOLDER}/eval_videos/')

for _ in range(10):

observations = env.reset()

red_obs = {}

blue_obs = {}

red_obs_ = {}

blue_obs_ = {}

for agent in red_agent_list:

red_obs[agent] = observations[agent]

for agent in blue_agent_list:

blue_obs[agent] = observations[agent]

observations = env.reset()

actions = {}

while not env.env_done:

actions = merge_dicts(red_team.choose_actions(red_obs), blue_team.choose_actions(blue_obs))

observations_, rewards, dones, _ = env.step(actions)

for agent in red_agent_list:

red_obs_[agent] = observations_[agent]

for agent in blue_agent_list:

blue_obs_[agent] = observations_[agent]

red_obs = red_obs_

blue_obs = blue_obs_