def worker(policy, sb, wl, seed, total_times, lm, model_path, da, bs, tc, ex):
env = TicTacToeXGameSpec(sb, wl, ex)
learn(policy,
env,
nsteps=sb * sb,
nstack=1,
seed=seed,
total_timesteps=total_times,
load_model=lm,
model_path=model_path,
data_augmentation=da,
BATCH_SIZE=bs,
TEMP_CTE=tc,
expert=ex)
policy_fn = CnnPolicy_slim_scope5x5_3_1x1
size_board = int(sys.argv[2])
winning_length = int(sys.argv[3])
TEMP_CTE = int(sys.argv[4])
BATCH_SIZE = int(sys.argv[5])
seed = int(sys.argv[6])
expert = int(sys.argv[7])
if expert == 1:
expert = True
elif expert == 2:
expert = False
model_path = ''
print('________________________policy', policy_fn, 'size_board',
size_board, 'winning_length', winning_length, 'batch size',
BATCH_SIZE, 'TEMP_CTE', TEMP_CTE, 'seed', seed, 'expert', expert)
env = TicTacToeXGameSpec(size_board, winning_length)
learn(policy_fn,
env,
nsteps=size_board * size_board,
nstack=1,
seed=seed,
total_timesteps=10000000,
load_model=False,
model_path=model_path,
data_augmentation=True,
BATCH_SIZE=BATCH_SIZE,
TEMP_CTE=TEMP_CTE,
expert=expert)
env.close()