def train(env_id, num_iters, seed, n=1, success_reward=1000, save_path='model/new_model'):
U.make_session(num_cpu=4).__enter__()
set_global_seeds(seed)
env = gym.make(env_id)
# env.update_adversary(n)
env=Monitor(env, log_dir, allow_early_resets=True)
env.seed(seed)
test_env = gym.make(env_id)
test_env.seed(seed)
gym.logger.setLevel(logging.WARN)
# debug not working
# num_cpu=4
# env=SubprocVecEnv([make_env(env_id,i,seed) for i in range(num_cpu)])
rew = PPO_RARL.learn(env, test_env, policy_fn,
timesteps_per_batch=2048,
clip_param=0.2, entcoeff=0.0,
optim_epochs=10, optim_stepsize=3e-4, optim_batchsize=64,
gamma=0.99, lam=0.95, schedule='constant', success_reward=success_reward,
save_path=save_path, max_iters=num_iters, callback=plot_callback
)
env.close()
return rew
def train(env_id, num_iters, seed, success_reward, save_path, q):
U.make_session(num_cpu=1).__enter__()
set_global_seeds(seed)
env = gym.make(env_id)
env=Monitor(env, log_dir, allow_early_resets=True)
test_env = gym.make(env_id)
env = bench.Monitor(env, logger.get_dir() and
osp.join(logger.get_dir(), "monitor.json"))
env.seed(seed)
test_env.seed(seed)
gym.logger.setLevel(logging.WARN)
rew = PPO.learn_with_human(env, test_env, policy_fn,
max_iters=num_iters,
timesteps_per_batch=2048,
clip_param=0.2, entcoeff=0.0,
optim_epochs=10, optim_stepsize=3e-4, optim_batchsize=64,
gamma=0.99, lam=0.95, schedule='constant', success_reward=success_reward,
save_path=save_path, callback=plot_callback, data_queue=q,
)
env.close()
return rew
