def test_main(tmp_path):
"""
tests for the logger module
"""
info("hi")
debug("shouldn't appear")
set_level(DEBUG)
debug("should appear")
configure(folder=str(tmp_path))
record("a", 3)
record("b", 2.5)
dump()
record("b", -2.5)
record("a", 5.5)
dump()
info("^^^ should see a = 5.5")
record_mean("b", -22.5)
record_mean("b", -44.4)
record("a", 5.5)
dump()
with ScopedConfigure(None, None):
info("^^^ should see b = 33.3")
with ScopedConfigure(str(tmp_path / "test-logger"), ["json"]):
record("b", -2.5)
dump()
reset()
record("a", "longasslongasslongasslongasslongasslongassvalue")
dump()
warn("hey")
error("oh")
record_dict({"test": 1})
def test_main(tmp_path):
"""
tests for the logger module
"""
info("hi")
debug("shouldn't appear")
assert get_level() == INFO
set_level(DEBUG)
assert get_level() == DEBUG
debug("should appear")
configure(folder=str(tmp_path))
assert get_dir() == str(tmp_path)
record("a", 3)
record("b", 2.5)
dump()
record("b", -2.5)
record("a", 5.5)
dump()
info("^^^ should see a = 5.5")
record("f", "this text \n \r should appear in one line")
dump()
info('^^^ should see f = "this text \n \r should appear in one line"')
record_mean("b", -22.5)
record_mean("b", -44.4)
record("a", 5.5)
dump()
with ScopedConfigure(None, None):
info("^^^ should see b = 33.3")
with ScopedConfigure(str(tmp_path / "test-logger"), ["json"]):
record("b", -2.5)
dump()
reset()
record("a", "longasslongasslongasslongasslongasslongassvalue")
dump()
warn("hey")
error("oh")
record_dict({"test": 1})
assert isinstance(get_log_dict(), dict) and set(get_log_dict().keys()) == {"test"}
def test_main():
"""
tests for the logger module
"""
info("hi")
debug("shouldn't appear")
set_level(DEBUG)
debug("should appear")
folder = "/tmp/testlogging"
if os.path.exists(folder):
shutil.rmtree(folder)
configure(folder=folder)
logkv("a", 3)
logkv("b", 2.5)
dumpkvs()
logkv("b", -2.5)
logkv("a", 5.5)
dumpkvs()
info("^^^ should see a = 5.5")
logkv_mean("b", -22.5)
logkv_mean("b", -44.4)
logkv("a", 5.5)
dumpkvs()
with ScopedConfigure(None, None):
info("^^^ should see b = 33.3")
with ScopedConfigure("/tmp/test-logger/", ["json"]):
logkv("b", -2.5)
dumpkvs()
reset()
logkv("a", "longasslongasslongasslongasslongasslongassvalue")
dumpkvs()
warn("hey")
error("oh")
logkvs({"test": 1})
def train(env_id, num_timesteps, seed):
# sess = util.single_threaded_session()
# sess.__enter__()
workerseed = seed + 10000 * MPI.COMM_WORLD.Get_rank()
# Create a new base directory like //home/marco/Reinforcement_Learning/Logs/openai-2018-05-21-12-27
log_dir = os.path.join(
energyplus_logbase_dir(),
datetime.datetime.now().strftime("openai-%Y-%m-%d-%H-%M"))
if not os.path.exists(log_dir + '/output'):
os.makedirs(log_dir + '/output')
os.environ["ENERGYPLUS_LOG"] = log_dir
model = os.getenv('ENERGYPLUS_MODEL')
if model is None:
print('Environment variable ENERGYPLUS_MODEL is not defined')
exit()
weather = os.getenv('ENERGYPLUS_WEATHER')
if weather is None:
print('Environment variable ENERGYPLUS_WEATHER is not defined')
exit()
# MPI is to parallelize training
# Logs the training in a file log.txt in the given directory
rank = MPI.COMM_WORLD.Get_rank()
if rank == 0:
print('train: init logger with dir={}'.format(log_dir)) # XXX
logger.configure(log_dir)
else:
logger.configure(format_strings=[])
logger.set_level(logger.DISABLED)
# Make Gym environment:
env = make_energyplus_env(env_id, workerseed)
###### EXPERIMENTS FROM FIRST PAPER: ###########################################
#
# trpo_mpi.learn(env, policy_fn,
# max_timesteps=num_timesteps,
# timesteps_per_batch=16*1024, max_kl=0.01, cg_iters=10, cg_damping=0.1,
# gamma=0.99, lam=0.98, vf_iters=5, vf_stepsize=1e-3)
# Apply TRPO algorithm from OpenAI baselines:
# action_noise = NormalActionNoise(mean=np.zeros(4), sigma=0.1 * np.ones(4))
#
# policy_kwargs_tqc = dict(n_critics=2, n_quantiles=25)
# model_tqc = TQC("MlpPolicy", env, top_quantiles_to_drop_per_net=2
# , verbose=1, policy_kwargs=policy_kwargs_tqc)
#
# model_ppo = PPO('MlpPolicy', env, verbose=1, n_steps=4096, batch_size=64, n_epochs=15)
# model_td3 = TD3('MlpPolicy', env, verbose=1, action_noise=action_noise)
# model_sac = SAC('MlpPolicy', env, verbose=1)
# model_ppolstm = PPO2(MlpLstmPolicy, env, verbose=1,n_steps=27, nminibatches=1)
#
# # Change the algorithm here:
#
# model_ppolstm.learn(total_timesteps=num_timesteps, log_interval=1, reset_num_timesteps=False)
# # model_ppo.learning_rate = 0
# # model_ppo.learn(total_timesteps=35040, reset_num_timesteps=False)
#
#####################################EXPERIMENTS 2: ###################################
sac_v2_lstm(env, num_timesteps, train=True, test=False)
#slac(env, num_timesteps)
env.close()