## Agent vybiraet dejstvie
# (candles=9(mb=>(2,4)?), tickers=4, trades=2)
# TODO actions for multy symbols market
action = agent.forward(observation)
## Execute action
observation, reward, done, info = market.step([action])
## Poluchaem otvet ot sredy
agent.backward(reward, terminal=done)
## Esli dostigli konca
if done:
observation = market.reset()
agent.reset_states()
done = False
log.info('Is terminal state. Reset..')
log.info('='*40)
log.info('Tick: {t} | {info}'.format(
t=tickcount, info=info
))
## Check point
if tickcount % 100 == 0:
agent.save_weights('{p}/dqn_{fn}_weights.h5f'.format(p=PATH, fn=ENV_NAME), overwrite=True)
## Time shift and counter
time.sleep(SLEEP)
tickcount += 1
