def test_list_files(self):
trade = Trade()
list = trade.list_tfdata_list()
print(list)
print('number of files->', trade.number_of_files)
def test_read_one_episode(self):
trade = Trade()
trade.new_episode()
num_of_records = 0
for i in range(1000):
if trade.new_sec():
self.assertIsNotNone(trade.boards)
self.assertIsNotNone(trade.buy_trade_price)
self.assertIsNotNone(trade.sell_trade_price)
self.assertIsNotNone(trade.buy_book_price)
self.assertIsNotNone(trade.sell_book_vol)
self.assertIsNotNone(trade.sell_book_price)
self.assertIsNotNone(trade.buy_book_vol)
self.assertIsNotNone(trade.market_sell_price)
self.assertIsNotNone(trade.buy_trade_vol)
self.assertIsNotNone(trade.fix_buy_price)
self.assertIsNotNone(trade.market_buy_price)
self.assertIsNotNone(trade.time)
num_of_records += 1
print(trade.time)
else:
break
print(num_of_records)
def test_read_one_episode(self):
trade = Trade()
trade.new_episode()
num_of_records = 0
while trade.new_sec():
self.assertIsNotNone(trade.board)
self.assertIsNotNone(trade.buy_trade_price)
self.assertIsNotNone(trade.sell_trade_price)
self.assertIsNotNone(trade.buy_book_price)
self.assertIsNotNone(trade.sell_book_vol)
self.assertIsNotNone(trade.sell_book_price)
self.assertIsNotNone(trade.buy_book_vol)
self.assertIsNotNone(trade.market_sell_price)
self.assertIsNotNone(trade.buy_trade_vol)
self.assertIsNotNone(trade.fix_buy_price)
self.assertIsNotNone(trade.market_buy_price)
self.assertIsNotNone(trade.time)
print(trade.time)
num_of_records += 1
print(num_of_records)
def test_new_episode(self):
trade = Trade()
trade.new_episode()
result = trade.new_sec()
print(result)
print(trade.time)
def one_episode(self, env: Trade):
s = env.new_episode()
while True:
action = self.policy(s)
print('action->', action, end='')
next_state, reward, done, info = env.step(action)
print('reward->', reward)
if done:
break
s = next_state
return reward
def setUp(self) -> None:
self.agent = Agent()
self.env = Trade()
self.trainer = Trainer(self.env, self.agent)
self.trainer.one_episode()
def test_buy_order_sell(self):
trade = Trade()
trade.new_episode()
print(trade.time)
trade.step(ACTION.BUY)
print(trade.time)
print(trade.buy_order_price)
print(trade.margin)
print(trade.time)
trade.step(ACTION.SELL)
print(trade.time)
print(trade.margin)
def test_env_buy(self):
trade = Trade()
trade.new_episode()
print(trade.time)
trade.step(ACTION.BUY)
print(trade.time)
def test_env_sell_buy(self):
trade = Trade()
trade.new_episode()
print(trade.time)
trade.step(ACTION.SELL_NOW)
print(trade.time)
print(trade.sell_order_price)
print(trade.margin)
# close transaction
print(trade.time)
trade.step(ACTION.BUY_NOW)
print(trade.time)
print(trade.margin)
estimated[i][e.a] = reward
if e.s.is_able_to_buy():
estimated[i][ACTION.BUY_NOW] = e.s.get_buy_now_reward()
else:
estimated[i][ACTION.BUY_NOW] = 0
estimated[i][ACTION.BUY] = 0
if e.s.is_able_to_sell():
estimated[i][ACTION.SELL_NOW] = e.s.get_sell_now_reward()
else:
estimated[i][ACTION.SELL_NOW] = 0
estimated[i][ACTION.SELL] = 0
loss = self.model.train_on_batch([states, rewards], estimated)
return loss
def update_model(self):
self.teacher_model.set_weights(self.model.get_weights())
if __name__ == '__main__':
trainer = Trainer()
env = Trade()
agent = Dqn()
# trainer.train(env, agent, eposode=100000, min_buffer_size=128)
trainer.train(env, agent, eposode=500000, min_buffer_size=15000)
def test_env_step(self):
trade = Trade()
trade.new_episode()
trade.step(ACTION.NOP)
trade.step(ACTION.BUY)
trade.step(ACTION.BUY_NOW)
trade.step(ACTION.SELL)
trade.step(ACTION.SELL_NOW)
def test_new_sec(self):
trade = Trade()
trade.new_episode()
trade.new_sec()
print(trade.time)
trade.new_sec()
print(trade.time)
trade.new_sec()
print(trade.time)
trade.new_sec()
print(trade.time)
def test_skip_count(self):
trade = Trade()
print(trade._skip_count())
def test_new_episode_files(self):
trade = Trade()
files = trade._new_episode_files()
print(files)
def test_actions(self):
trade = Trade()
trade.new_episode()
print(trade.action_space)
def test_tfrecords(self):
trade = Trade()
trade._new_episode(0, 2)
for data in trade.dataset:
trade.decode_dataset(data)
print(trade.time)
break
for data in trade.dataset:
trade.decode_dataset(data)
print(trade.time)
break
trade._new_episode(0, 2)
for data in trade.dataset:
trade.decode_dataset(data)
print(trade.time)
break