def __init__(self):
'''
应用初始化
'''
logger.init_logger(CONFIG.LOG_CONF, suffix=options.port)
settings = {
'static_path':
os.path.join(options.project_path, 'static'),
'template_path':
os.path.join(options.project_path, 'tpl'),
'xsrf_cookies':
False,
'site_title':
'demo',
'debug':
options.debug,
'ui_modules':
uimodules,
'ui_methods':
uimethods,
"session_mgr":
TornadoSessionManager(CONFIG.COOKIE_NAME, CONFIG.COOKIE_SECRET,
MClient(CONFIG.MC_SERVERS)),
}
handlers = [
(r"/", IndexHandler),
]
tornado.web.Application.__init__(self, handlers, **settings)
def __init__(self,
model: BaseRLModel = PPO2,
policy: BasePolicy = MlpLnLstmPolicy,
reward_strategy: BaseRewardStrategy = IncrementalProfit,
exchange_args: Dict = {},
**kwargs):
self.logger = kwargs.get(
'logger',
init_logger(__name__, show_debug=kwargs.get('show_debug', True)))
self.Model = model
self.Policy = policy
self.Reward_Strategy = reward_strategy
self.exchange_args = exchange_args
self.tensorboard_path = kwargs.get('tensorboard_path', None)
self.input_data_path = kwargs.get('input_data_path',
'data/input/EURUSD60.csv')
self.params_db_path = kwargs.get('params_db_path',
'sqlite:///data/params.db')
self.date_format = kwargs.get('date_format',
ProviderDateFormat.DATETIME_HOUR_24)
self.model_verbose = kwargs.get('model_verbose', 1)
self.n_envs = kwargs.get('n_envs', os.cpu_count())
self.n_minibatches = kwargs.get('n_minibatches', self.n_envs)
self.train_split_percentage = kwargs.get('train_split_percentage', 0.8)
self.data_provider = kwargs.get('data_provider', 'static')
self.initialize_data()
self.initialize_optuna()
self.logger.debug(f'Initialize RLTrader: {self.study_name}')
def __init__(self,
modelClass: BaseRLModel = PPO2,
policyClass: BasePolicy = MlpPolicy,
**kwargs):
self.logger = init_logger(__name__,
show_debug=kwargs.get('show_debug', True))
self.Model = modelClass
self.Policy = policyClass
self.tensorboard_path = kwargs.get('tensorboard_path', None)
self.input_data_path = kwargs.get('input_data_path', None)
self.params_db_path = kwargs.get('params_db_path',
'sqlite:///data/params.db')
self.date_format = kwargs.get('date_format',
ProviderDateFormat.DATETIME_HOUR_12)
self.model_verbose = kwargs.get('model_verbose', 1)
self.nminibatches = kwargs.get('nminibatches', 1)
self.train_split_percentage = kwargs.get('train_split_percentage', 0.8)
self.initialize_data()
self.initialize_optuna()
self.logger.debug(f'Initialize RLTrader: {self.study_name}')
def __init__(self,
data_provider: BaseDataProvider,
initial_balance=10000,
commission=0.0025,
**kwargs):
super(TradingEnv, self).__init__()
self.logger = init_logger(__name__,
show_debug=kwargs.get('show_debug', True))
self.data_provider = data_provider
self.initial_balance = initial_balance
self.commission = commission
self.reward_fn = kwargs.get('reward_fn',
self._reward_incremental_profit)
self.benchmarks = kwargs.get('benchmarks', [])
self.enable_stationarization = kwargs.get('enable_stationarization',
True)
self.action_space = spaces.Discrete(3)
n_features = 5 + len(self.data_provider.columns)
self.obs_shape = (1, n_features)
self.observation_space = spaces.Box(low=0,
high=1,
shape=self.obs_shape,
dtype=np.float16)
self.observations = pd.DataFrame(None,
columns=self.data_provider.columns)
def __init__(self,
data_provider: BaseDataProvider,
reward_strategy: BaseRewardStrategy = IncrementalProfit,
trade_strategy: BaseTradeStrategy = SimulatedTradeStrategy,
initial_balance: int = 10000,
commissionPercent: float = 0.25,
maxSlippagePercent: float = 2.0,
**kwargs):
super(TradingEnv, self).__init__()
self.logger = kwargs.get('logger', init_logger(__name__, show_debug=kwargs.get('show_debug', True)))
self.base_precision: int = kwargs.get('base_precision', 2)
self.asset_precision: int = kwargs.get('asset_precision', 8)
self.min_cost_limit: float = kwargs.get('min_cost_limit', 1E-3)
self.min_amount_limit: float = kwargs.get('min_amount_limit', 1E-3)
self.initial_balance = round(initial_balance, self.base_precision)
self.commissionPercent = commissionPercent
self.maxSlippagePercent = maxSlippagePercent
self.data_provider = data_provider
self.reward_strategy = reward_strategy()
self.trade_strategy = trade_strategy(commissionPercent=self.commissionPercent,
maxSlippagePercent=self.maxSlippagePercent,
base_precision=self.base_precision,
asset_precision=self.asset_precision,
min_cost_limit=self.min_cost_limit,
min_amount_limit=self.min_amount_limit)
self.render_benchmarks: List[Dict] = kwargs.get('render_benchmarks', [])
self.render_benchmarks = [
{
'label': 'Buy and HODL',
'values': buy_and_hodl(self.data_provider.data_frame['Close'], initial_balance, commissionPercent)
},
{
'label': 'RSI Divergence',
'values': rsi_divergence(self.data_provider.data_frame['Close'], initial_balance, commissionPercent)
},
{
'label': 'SMA Crossover',
'values': sma_crossover(self.data_provider.data_frame['Close'], initial_balance, commissionPercent)
},
]
self.normalize_obs: bool = kwargs.get('normalize_obs', True)
self.stationarize_obs: bool = kwargs.get('stationarize_obs', True)
self.normalize_rewards: bool = kwargs.get('normalize_rewards', False)
self.stationarize_rewards: bool = kwargs.get('stationarize_rewards', True)
self.n_discrete_actions: int = kwargs.get('n_discrete_actions', 24)
self.action_space = spaces.Discrete(self.n_discrete_actions)
self.n_features = 6 + len(self.data_provider.columns)
self.obs_shape = (1, self.n_features)
self.observation_space = spaces.Box(low=0, high=1, shape=self.obs_shape, dtype=np.float16)
self.observations = pd.DataFrame(None, columns=self.data_provider.columns)
def __init__(self,
data_provider: BaseDataProvider,
reward_strategy: BaseRewardStrategy = IncrementalProfit,
trade_strategy: BaseTradeStrategy = SimulatedTradeStrategy,
initial_balance: int = 10000,
commissionPercent: float = 0.25,
maxSlippagePercent: float = 2.0,
**kwargs):
super(TradingEnv, self).__init__()
self.logger = kwargs.get(
'logger',
init_logger(__name__, show_debug=kwargs.get('show_debug', True)))
self.base_precision: int = kwargs.get('base_precision', 2)
self.asset_precision: int = kwargs.get('asset_precision', 8)
self.min_cost_limit: float = kwargs.get('min_cost_limit', 1E-3)
self.min_amount_limit: float = kwargs.get('min_amount_limit', 1E-3)
self.initial_balance = round(initial_balance, self.base_precision)
self.commissionPercent = commissionPercent
self.maxSlippagePercent = maxSlippagePercent
self.data_provider = data_provider
self.reward_strategy = reward_strategy()
self.trade_strategy = trade_strategy(
commissionPercent=self.commissionPercent,
maxSlippagePercent=self.maxSlippagePercent,
base_precision=self.base_precision,
asset_precision=self.asset_precision,
min_cost_limit=self.min_cost_limit,
min_amount_limit=self.min_amount_limit)
self.render_benchmarks: List[Dict] = kwargs.get(
'render_benchmarks', [])
self.normalize_obs: bool = kwargs.get('normalize_obs', True)
self.stationarize_obs: bool = kwargs.get('stationarize_obs', True)
self.normalize_rewards: bool = kwargs.get('normalize_rewards', False)
self.stationarize_rewards: bool = kwargs.get('stationarize_rewards',
True)
self.n_discrete_actions: int = kwargs.get('n_discrete_actions', 24)
self.action_space = spaces.Discrete(self.n_discrete_actions)
self.n_features = 5 + len(self.data_provider.columns)
self.obs_shape = (1, self.n_features)
self.observation_space = spaces.Box(low=0,
high=1,
shape=self.obs_shape,
dtype=np.float16)
self.observations = pd.DataFrame(None,
columns=self.data_provider.columns)
self.fast_transform_observations = FastTransform()
self.fast_transform_account_history = FastTransform()
self.fast_log_observation = self.observations
def __init__(self, file_name: str = "database.db"):
# Test if 'data' folder exists. If not, create it.
directory = "data"
self.create_directory_if_not_exists(directory)
# All database files will be located at data folder
self.conn = sqlite3.connect(f"{directory}/{file_name}")
self.cursor = self.conn.cursor()
self.logger = init_logger(self.__class__.__name__)
def __init__(self):
'''
应用初始化
'''
logger.init_logger(CONFIG.LOG_CONF, suffix=options.port)
settings = {
'static_path': os.path.join(options.project_path, 'static'),
'template_path': os.path.join(options.project_path, 'tpl'),
'xsrf_cookies': False,
'site_title': 'demo',
'debug': options.debug,
'ui_modules':uimodules,
'ui_methods':uimethods,
"session_mgr": TornadoSessionManager(CONFIG.COOKIE_NAME,
CONFIG.COOKIE_SECRET,
MClient(CONFIG.MC_SERVERS)),
}
handlers = [
(r"/", IndexHandler),
]
tornado.web.Application.__init__(self, handlers, **settings)
def __init__(
self,
model_actor: BaseRLModel = PPO2,
policy:
BasePolicy = MlpLnLstmPolicy #(ActorCriticPolicy or str) The policy model to use (MlpPolicy, CnnPolicy, CnnLstmPolicy, ...)
,
reward_strategy: Reward_Strategy_BASE = RewardPnL #IncrementalProfit
,
exchange_args: Dict = {},
**kwargs):
self.model_actor: BaseRLModel = model_actor
self.policy = policy
self.Reward_Strategy = reward_strategy
self.exchange_args = exchange_args
self.logger = kwargs.get(
'logger',
init_logger(__name__, show_debug=kwargs.get('show_debug', True)))
self.db_path = kwargs.get('db_path', 'sqlite:///data/params.db')
self.date_format = kwargs.get('date_format',
ProviderDateFormat.DATETIME_HOUR_24)
self.data_path = kwargs.get('data_path',
'data/input/coinbase-1h-btc-usd.csv')
self.data_train_split_pct = kwargs.get('train_split_percentage', 0.8)
self.data_provider = kwargs.get('data_provider', 'static')
#self.columns_map = kwargs.get('columns_map', {})
self.n_envs = kwargs.get('n_envs', os.cpu_count())
self.n_minibatches = kwargs.get('n_minibatches', self.n_envs)
self.model_logs_tb = kwargs.get('tensorboard_path',
os.path.join('data', 'logs_tb'))
self.model_verbose = kwargs.get('model_verbose', 1)
self.do_load_raw_data: bool = kwargs.get('do_load_raw_data', True)
self.features_to_add: str = kwargs.get('features_to_add', 'none')
self.initialize_data(self.do_load_raw_data, self.features_to_add)
self.initialize_db_optuna() #optimization for hyper param search
self.logger.info(
f'sucsessfully Initialize RLTrader study name {self.study_name} , open terminal, tensorboard --logdir={self.model_logs_tb}, click to http://localhost:6006/'
)
import numpy as np
from lib.RLTrader import RLTrader
from lib.TraderArgs import TraderArgs
from lib.util.logger import init_logger
np.warnings.filterwarnings('ignore')
option_parser = TraderArgs()
args = option_parser.get_args()
if __name__ == '__main__':
logger = init_logger(__name__, show_debug=args.debug)
trader = RLTrader(**vars(args), logger=logger)
if args.command == 'optimize':
trader.optimize(n_trials=args.trials,
n_parallel_jobs=args.parallel_jobs)
elif args.command == 'train':
trader.train(n_epochs=args.epochs)
elif args.command == 'test':
trader.test(model_epoch=args.model_epoch, should_render=args.no_render)
elif args.command == 'opt-train-test':
trader.optimize(args.trials, args.parallel_jobs)
trader.train(n_epochs=args.train_epochs,
test_trained_model=args.no_test,
render_trained_model=args.no_render)
