def load(size, seed=0): """Load a deep sea experiment with the prescribed settings.""" env = deep_sea.DeepSea( size=size, deterministic=False, seed=seed, ) env.bsuite_num_episodes = sweep.NUM_EPISODES return env
def make_object_under_test(self): return deep_sea.DeepSea(5, deterministic=False)
def make_object_under_test(self): return deep_sea.DeepSea(10)
from numpy import random import sys import bsuite from bsuite.logging.csv_logging import wrap_environment from bsuite.experiments.deep_sea import deep_sea from wrap2 import derived from DeepseaEncoder import DeepSea_encoder #Initializing params total_steps = 1e5 / 2 n_size = 20 LAST = 2 SAVE_PATH_RAND = 'results/deepsea/TS20-LONG' env = deep_sea.DeepSea(n_size, seed=int(sys.argv[1])) logname = 'deep_sea/' + sys.argv[1] env = wrap_environment(env, logname, SAVE_PATH_RAND, overwrite=True) discount = 0.95 n_actions = 2 ql = derived(n_size * n_size, n_actions, discount, 2, 5, 40) #cross-validated params: L20:(2,5,40) ; L10:(4,10,20) encoder = DeepSea_encoder(n_size) #Interacting with environment num_trials = 1 counter = 0 episodic_reward = [] n_episodes = 0 for i in range(num_trials): total_reward = 0 prev_act = 0