def train():
""" Trains an RL model.
First initializes environment, logging, and machine learning model. Then iterates
through epochs of training and prints score intermittently.
"""
train = {}
train['env'] = gym.make(params['env_name'])
train['env'].init(params)
train['model'] = PPO(params, train['env'].observation_space.shape[0]).to(params['device'])
if params['transfer']:
train['model'].load_state_dict(torch.load(sys.argv[1]))
logger = Logger()
score = 0.0
for n_epi in range(10**6):
ep_score = episode(train, n_epi)
logger.episode_score(ep_score, n_epi)
score += ep_score
if n_epi % params['print_interval'] == 0 and n_epi != 0:
print(f"Episode #{n_epi:5d} | Avg Score : {score / params['print_interval']:2.2f}")
if n_epi >= 0:
logger.save_model(score/params['print_interval'], train['model'].state_dict(), n_epi)
score = 0.0
train['model'].train_net()
env.close()
