def create_app():
_app = Agent(pull_interval=5, debug=True)
return _app
import tensorflow as tf from env.main import Account from agent.main import Agent, Access, Framework env = Account() init = env.reset() print(init.shape) name = 'W0' input_shape = [381, 240, 58] action_size = 3 A = Access(input_shape, action_size) W0 = Agent(name, A, input_shape, action_size) print(W0.a_interface) # with tf.Session() as sess: # sess.run(tf.global_variables_initializer()) # W0.init_or_update_local(sess) # da = W0.get_deterministic_policy_action(sess, init) # sa = W0.get_stochastic_action(sess, init) # # next_state, reward, done = env.step(sa) # print(next_state.shape) # print(reward) # print(done)
# while True:
# action = agent.get_stochastic_policy(sess, state)
# next_state, reward, done = env.step(action)
# agent.update_cache(state, action, reward, next_state, done)
# state = next_state
# if done:
# break
#
# agent.update_value_net(sess)
# agent.update_target_net(sess)
env = Account()
state = env.reset()
agent = Agent()
while True:
action = agent.get_stochastic_policy(state)
next_state, reward, done = env.step(action)
agent.update_cache(state, action, reward, next_state, done)
state = next_state
if done:
break
agent.update_target()
agent.update_eval()
agent.save_model()
agent.restore_model()
import gc
import time
import numpy as np
from visdom import Visdom
from agent.main import Agent
from emulator.main import Account
# viz = Visdom()
# assert viz.check_connection()
env = Account()
state = env.reset()
image_shape = state.shape
print(image_shape)
agent = Agent(image_shape, 3)
max_episodes = 1000
global_step = 0
for episode in range(max_episodes):
state = env.reset()
cache_reward = []
cache_valaue = []
while True:
global_step += 1
action = agent.get_epsilon_policy(np.expand_dims(state, 0))
next_state, reward, done = env.step(action)
agent.append(state, action, reward, done, next_state)
cache_reward.append(reward)
cache_valaue.append(env.total_value)
def create_app():
_app = Agent(pull_interval=5)
return _app
import numpy as np
import pandas as pd
import seaborn as sns
import tensorflow as tf
from agent.main import Access, Agent
from env.main import Account
tf.reset_default_graph()
sns.set_style('whitegrid')
%matplotlib
inputs_shape = [381, 240, 58]
action_size = 3
config = tf.ConfigProto()
config.gpu_options.allow_growth = True
with tf.Session(config=config) as sess:
with tf.device("/cpu:0"):
A = Access(inputs_shape, action_size)
W = Agent('W0', A, inputs_shape, action_size)
A.restore(sess,'model/saver_1.ckpt')
W.init_or_update_local(sess)
env = Account()
state = env.reset()
for _ in range(466):
action = W.get_deterministic_policy_action(sess, state)
state, reward, done = env.step(action)