In [1]:

    

import tensorflow as tf
from emulator.main import Account
from agent.access import Access
from agent.main import Framework
import numpy as np
import pandas as pd
import seaborn as sns
%matplotlib inline


    

In [2]:

    

state_size = [50, 58, 5]
action_size = 3

with tf.Session() as sess:
    with tf.device("/cpu:0"):
        A = Access(state_size, action_size)
        W = Framework(A, state_size, action_size, 'AC')
        A.restore(sess,'model/saver_1.ckpt')
        W.init_network(sess)
        env = Account()
        state = env.reset()
        net_value = []
        reward_list = []
        for _ in range(1400):
            action = W.get_deterministic_policy_action(sess, state)
            reward, state, done = env.step(action)
            reward_list.append(reward)
            net_value.append(env.total_value)


    

    




graph AC
INFO:tensorflow:Restoring parameters from model/saver_1.ckpt

In [3]:

    

pd.Series(net_value).plot(figsize=(16,6))


    

    
Out[3]:





<matplotlib.axes._subplots.AxesSubplot at 0x2048ec89f60>






    

In [4]:

    

pd.Series(net_value[1000:]).plot(figsize=(16,6))


    

    
Out[4]:





<matplotlib.axes._subplots.AxesSubplot at 0x2048f397cc0>






    

In [5]:

    

pd.Series(reward_list).plot(figsize=(16,6))
pd.Series(np.zeros_like(reward_list)).plot(figsize=(16,6), color='r')


    

    
Out[5]:





<matplotlib.axes._subplots.AxesSubplot at 0x2048ed7c550>






    

In [6]:

    

reward_list = reward_list[1000:]
pd.Series(reward_list).plot(figsize=(16,6))
pd.Series(np.zeros_like(reward_list)).plot(figsize=(16,6), color='r')


    

    
Out[6]:





<matplotlib.axes._subplots.AxesSubplot at 0x2048ef18550>






    

In [ ]: