class TestEnv(TestCase):
def test_sell(self):
self.env = Env(
balance=250000,
FX_DATA_FILE='../data/raw/FX_Demo/sample_USD_JPY_S5.pickle')
sell_mount = self.env.step(action='sell', mount=1)
assert sell_mount == {'fail': 0}
self.env.step(action='buy', mount=1)
sell_mount = self.env.step(action='sell', mount=1)
assert sell_mount == {'success': 249999.98500000002}
assert self.env.stock_balance == 0
assert self.env.balance == 249999.98500000002
def test_buy(self):
self.env = Env(
balance=250000,
FX_DATA_FILE='../data/raw/FX_Demo/sample_USD_JPY_S5.pickle')
buy_mount = self.env.step(action='buy', mount=1)
assert buy_mount == {'success': 249999.987}
assert self.env.stock_balance == 1
assert self.env.balance == 249888.668
def test_stay(self):
self.env = Env(
balance=250000,
FX_DATA_FILE='../data/raw/FX_Demo/sample_USD_JPY_S5.pickle')
stay_mount = self.env.step(action='stay', mount=1)
assert stay_mount == {'success': 250000.0}
assert self.env.stock_balance == 0
assert self.env.balance == 250000
def test_reset(self):
self.env = Env(
balance=250000,
FX_DATA_FILE='../data/raw/FX_Demo/sample_USD_JPY_S5.pickle')
self.env.step(action='buy', mount=1)
self.env.step(action='sell', mount=1)
self.env.reset()
assert self.env.stock_balance == 0
assert self.env.balance == 250000
if __name__ == "__main__":
env = Env()
agent = TestAgent(action_size=6, env=env)
agent.load_model("./save_model/supermario_per.h5")
for e in range(EPISODES):
done = False
max_x = 0
now_x = 0
hold_frame = 0
before_max_x = 200
start_position = 500
step, score = 0, 0
observe = env.reset(start_position=start_position)
state = pre_processing(observe)
history = np.stack((state, state, state, state), axis=2)
history = np.reshape([history], (1, 84, 84, 4))
action_count = 0
real_action, action = 0, 0
while not done:
step += 1
action = agent.get_action(history)
if action == 0:
real_action = 0
elif action == 1: