def _delete(instance, client, status, force_delete=False):
instance = client.instances.get(instance)
change_status(instance, client, status)
if not force_delete:
client.instances.delete(instance)
else:
client.instances.force_delete(instance)
def _attach_as_admin(self, status):
client = self.clients("trove")
manager = client.instances
instance, conf = manager.get(self.context['admin_instance']
), self.context['admin_configuration']
change_status(instance, client, status)
manager.modify(instance, conf)
def predict_func(args, log_dir):
# Required: testx, model_dir, weight
id = args.id
csv_header = args.header
change_status('loading', id)
test_data = load_file(args.testx, csv_header)
model = load_model(args.weight)
change_status('executing', id)
result = model.predict(test_data)
model_dir = args.model
model_file = 'preprocessed/result.json'
model_path = os.path.join(model_dir, model_file)
with open(model_path) as f:
model_parser = json.load(f)
for key, value in model_parser['layers'].items():
if value['type'] == 'Output':
loss = value['params']['loss']
if 'entropy' in loss:
problem = 'classification'
else:
problem = 'regression'
with open(os.path.join(log_dir, 'type'), 'w') as f:
f.write(str(problem))
with open(os.path.join(log_dir, 'result'), 'w') as f:
for i in result:
f.write(','.join(map(str, i)) + '\n')
def _resize(instance, client, status, size):
instance = client.instances.get(instance)
change_status(instance, client, status)
client.instances.resize_volume(instance['id'], size)
utils.wait_for(instance,
update_resource=client.instances.get,
ready_statuse=["RESIZE"])
check_ready(instance)
def _resize(instance, client, status, flavor_id):
instance = client.instances.get(instance)
change_status(instance, client, status)
client.instances.resize_instance(instance, flavor_id)
utils.wait_for(instance,
update_resource=client.instances.get,
ready_statuse=["RESIZE"])
check_ready(instance)
def _restart(instance, client, status):
instance = client.instances.get(instance)
change_status(
instance, status,
client.configurations if status == "RESTART_REQUIRED" else None)
instance.restart
utils.wait_for(instance,
ready_statuse=["REBOOT"],
update_resource=client.instances.get)
check_ready(instance)
def validate_func(args, log_dir):
# Required: testx, testy, model_dir, weight
id = args.id
csv_header = args.header
change_status('loading', id)
test_data = load_file(args.testx, csv_header)
test_target = load_file(args.testy, csv_header)
model = load_model(args.weight)
change_status('executing', id)
result = model.predict(test_data)
model_dir = args.model
model_file = 'preprocessed/result.json'
model_path = os.path.join(model_dir, model_file)
with open(model_path) as f:
model_parser = json.load(f)
for key, value in model_parser['layers'].items():
if value['type'] == 'Output':
loss = value['params']['loss']
if 'entropy' in loss:
problem = 'classification'
else:
problem = 'regression'
with open(os.path.join(log_dir, 'type'), 'w') as f:
f.write(str(problem))
with open(os.path.join(log_dir, 'result'), 'w') as f:
for i in result:
f.write(','.join(map(str, i)) + '\n')
if problem == 'classification':
classes = len(result[0])
conf_matrix = [[0.0] * classes] * classes
total_count = len(result)
score = (1 / float(total_count))
for i in range(len(result)):
conf_matrix[test_target[i].argmax()][result[i].argmax()] += score
with open(os.path.join(log_dir, 'evaluate'), 'w') as f:
for i in conf_matrix:
f.write(','.join(map(str, i)) + '\n')
def get_data_allocate(strategy, data_type):
dd = Calendar.today()
a = Calendar.in_business(dd, day=True)
if (a == False):
print('非工作日！')
exit(0)
t1 = time.clock()
print(time.strftime('%Y-%m-%d %H:%M:%S', time.localtime(time.time())))
command = "python {}\daily_{}_{}.py".format(project_dir, data_type,
strategy)
list = []
if (strategy == 'min5'):
list = get_min5_time()
if (strategy == 'min15'):
list = get_min15_time()
if (strategy == 'min30'):
list = get_min30_time()
if (strategy == 'min60'):
list = get_min60_time()
if (strategy == 'day'):
list = get_day_time()
print(strategy)
while True:
# if in_business(datetime.now(), strategy, True):
x = datetime.now().hour * 60 + datetime.now().minute
# print ( datetime.now().hour,datetime.now().minute)
# print(x)
# if(x>list[-1]):
if (x > list[-1]):
print(list[-1])
print('收盘！')
exit(0)
if (x == 11 * 60 + 45):
print('上午结束')
exit(0)
if (x in list):
print('finded')
t1 = time.clock()
subprocess.call(command, shell=True)
t3 = time.strftime('%Y-%m-%d %H:%M:%S',
time.localtime(time.time()))
print(t3)
dit = {'date': t3, 'status': 1, 'kline': strategy, 'other': ''}
model_list['kline_data_update_mark'].insert(dit)
t2 = time.clock()
a = (change_status(strategy)['time']) - ((t2 - t1) / 60)
print(a)
if (a > 0):
time.sleep(a * 60)
print('睡眠结束')
else:
print('finding:')
time.sleep(10)
def train_func(args, log_dir):
id = args.id
change_status('loading', id)
model, model_dir, (trainx, trainy, testx, testy) = build_model(args)
# Callback_1
history_callback = Batch_History()
# Callback_2
state_file = os.path.join(model_dir, 'state.json')
#state_file = "/home/plash/petpen/state.json"
state_callback = Model_state(state_file, model.config)
# Callback_3
rl_callback = RealtimeLogger(os.path.join(log_dir, 'realtime_logging.txt'))
change_status('running', id)
history = model.train(
callbacks=[history_callback, state_callback, rl_callback])
save_history(os.path.join(log_dir, 'train_log'), history, history_callback)
model_result_path = os.path.dirname(log_dir)
model.save(os.path.join(model_result_path, 'weights.h5'))
def _create(instance, client, status, **kwargs):
b_manager, i_manager = client.backups, client.instances
change_status(instance, client, status)
backup = b_manager.create(instance=instance['id'], **kwargs)
check_ready(instance['id'], i_manager)
b_manager.delete(backup)
if not os.path.exists(model_result_path):
os.mkdir(model_result_path)
log_dir = os.path.join(model_result_path, 'logs')
os.mkdir(log_dir)
error_log_file = os.path.join(log_dir, 'error_log')
id = args.id
try:
if 'train' in args.func.__name__:
args.func(args, log_dir)
else:
args.func(args, model_result_path)
change_status('finish', id)
except:
exc_type, exc_value, exc_traceback = sys.exc_info()
lines = traceback.format_exception(exc_type, exc_value, exc_traceback)
with open(error_log_file, 'w') as error_log:
for line in lines:
error_log.write(line)
with open(os.path.join(model_dir, 'state.json'), 'w') as state_file:
info = {'status': 'error', 'error_log_file': error_log_file}
json.dump(info, state_file)
change_status('error', id)