def fn_opt(trial: optuna.Trial) -> int:
try:
net_arch = trial.suggest_categorical('net_arch',
['CnnPolicy', 'LnCnnPolicy'])
gamma = trial.suggest_categorical(
'gamma', [0.9, 0.95, 0.98, 0.99, 0.995, 0.999, 0.9999])
learning_rate = trial.suggest_loguniform('lr', 1e-5, 1)
batch_size = trial.suggest_categorical('batch_size',
[16, 32, 64, 128, 256, 512])
buffer_size = trial.suggest_categorical(
'buffer_size', [int(1e4), int(1e5), int(1e6)])
learning_starts = trial.suggest_categorical('learning_starts',
[0, 25, 50, 75, 100])
gradient_steps = trial.suggest_categorical('gradient_steps',
[5, 15, 20, 50, 100, 300])
ent_coef = trial.suggest_categorical(
'ent_coef', ['auto', 0.5, 0.1, 0.05, 0.01, 0.0001])
target_entropy = 'auto'
if ent_coef == 'auto':
target_entropy = trial.suggest_categorical(
'target_entropy', ['auto', -1, -10, -20, -50, -100])
env = fn_gym.FNGym(0.2)
model = fn_sac.FNSAC(net_arch,
env,
gamma=gamma,
learning_rate=learning_rate,
batch_size=batch_size,
buffer_size=buffer_size,
learning_starts=learning_starts,
gradient_steps=gradient_steps,
ent_coef=ent_coef,
target_entropy=target_entropy)
for train_count in range(10):
model.learn(total_timesteps=200)
trial.report(env.get_running_reward(), (train_count + 1) * 200)
if trial.should_prune():
raise optuna.TrialPruned()
except KeyboardInterrupt:
input('Keyboard Interrupt. Press any key to continue')
raise ValueError("Exit Trial, Keyboard Interrupt")
return env.get_running_reward()
import fnai.fn_gym as fn_gym
from stable_baselines import SAC
from stable_baselines.common.policies import CnnLnLstmPolicy
from stable_baselines import PPO2
from os import path
parser = argparse.ArgumentParser(description='Train the model')
parser.add_argument(nargs='?', default='none', dest='model_path',
help='(optional) parameters to start training from')
args = parser.parse_args()
env = fn_gym.FNGym(0.2)
print(env.observation_space)
print(env.action_space)
model = PPO2(CnnLnLstmPolicy, env, tensorboard_log='sac_fn_tensorboard', nminibatches=1)
if path.exists(args.model_path):
print(f'Loading Model: {args.model_path}')
model.load(args.model_path)
else:
print('Using new model')
keep_training=True
log_num = 0
manual_log_num = 0
train_interval = 100000
while keep_training:
import fnai.fn_gym as fn_gym
import time
import numpy as np
from PIL import Image
test_gym = fn_gym.FNGym(0.2)
init_img = test_gym.reset()
for i in range(500):
screenshot = test_gym.d3d_buff.screenshot(region=test_gym.win_coords)
print(screenshot[25, 10, 0])
"""
print(np.shape(init_img))
obs, rew, done, info = test_gym.step([1.0, 1.0, 1.0])
print(np.shape(obs))
Image.fromarray(obs).save('test.png')
"""
import stable_baselines.common.env_checker as env_checker import fnai.fn_gym as fn_gym test_gym = fn_gym.FNGym(0.2, True) env_checker.check_env(test_gym, warn=True, skip_render_check=True)