def actions():
state_map = request.args.get('state_map')
side = request.args.get('side')
if not state_map or side not in ('b', 'r'):
return json.dumps({'error': True, 'msg': 'invalid params', 'data': {'state_map': state_map, 'side': side}})
result = KoreanChess.get_actions(json.loads(state_map), side)
return json.dumps(result)
def action():
state_map = request.args.get('state_map')
side = request.args.get('side')
if not state_map or side not in ('b', 'r'):
return json.dumps({
'error': True,
'msg': 'invalid params',
'data': {
'state_map': state_map,
'side': side
}
})
state_map = json.loads(state_map)
filter_state_map(state_map)
if side == 'b':
reverse_state_map = KoreanChess.reverse_state_map(state_map)
db_name = './q_blue.db'
else:
reverse_state_map = state_map
db_name = './q_red.db'
state_key = KoreanChess.convert_state_key(reverse_state_map)
conn = sqlite3.connect(db_name)
c = conn.cursor()
c.execute("SELECT quality_json FROM t_quality WHERE state_key='" +
state_key + "'")
result = c.fetchone()
reversed_state_map = KoreanChess.reverse_state_map(state_map)
actions = KoreanChess.get_actions(reversed_state_map, side)
if result:
if result[0] == '0':
action = KoreanChess.similar_action(actions, state_key, side, c)
else:
q_values = json.loads(result[0])
max_action = int(
max(q_values.iteritems(), key=operator.itemgetter(1))[0])
if len(actions) <= max_action:
action = KoreanChess.similar_action(actions, state_key, side,
c)
else:
action = actions[max_action]
else:
action = KoreanChess.similar_action(actions, state_key, side, c)
c.close()
return json.dumps(action)
j = 0
while not blue_done and not red_done:
if j >= 400:
break
blue_action = env.get_action_es(blue_state, kcu.BLUE)
red_state, blue_reward, blue_done, is_draw = env.step(blue_action, blue_state)
env.print_map(red_state, kcu.RED, i, j, blue_reward_all, red_reward_all, kcu.BLUE if blue_done else False,
is_draw, blue_win_cnt, red_win_cnt)
if blue_action is False and red_action is False:
print('there is no action.')
break
if old_red_state:
cur_reward = (red_reward - blue_reward * .9)
red_q = KoreanChess.get_q_from_es(red_state, kcu.RED)
max_action = int(max(red_q.iteritems(), key=operator.itemgetter(1))[1])
if red_q:
cur_q_value = cur_reward + dis * max_action
else:
cur_q_value = cur_reward
KoreanChess.update_es_q(old_red_state, red_action, cur_q_value)
red_reward_all += (red_reward - blue_reward * .9)
if blue_done and not is_draw:
blue_win_cnt += 1
break
red_action = env.get_action(Q_red, red_state, i, True)
key = line.strip()
else:
q_value = np.array(json.loads(line.strip()))
if np.sum(q_value) == 0:
quality_json = '0'
else:
quality_dict = {}
for j, q in enumerate(q_value):
if q != 0:
quality_dict[j] = q
quality_json = json.dumps(quality_dict)
# Insert a row of data
c.execute(
"INSERT INTO t_quality VALUES " + "('%s', '%s', %d, '%s', %d)" %
((KoreanChess.compress_state_key(key), quality_json, len(q_value),
datetime.datetime.now().strftime('%Y-%m-%d %H:%I:%S'), 1)))
i += 1
q_file.close()
# Save (commit) the changes
conn.commit()
# We can also close the connection if we are done with it.
# Just be sure any changes have been committed or they will be lost.
conn.close()
conn = sqlite3.connect('./q_red.db')
c = conn.cursor()
from elasticsearch import Elasticsearch as ES
from env.korean_chess import KoreanChess
import sys
es = ES('52.79.135.2:80')
q_file = open('./q_blue_with_data.txt')
i = 0
key = None
bulk_list = []
blue_index = {"index": {"_index": "i_irelia_state", "_type": "t_blue_state"}}
for line in q_file:
if i % 2 is 0:
key = line.strip()
decomp_key = KoreanChess.decompress_state_key(key)
bulk_list.append(blue_index)
bulk_list.append({"state": decomp_key})
if i % 1000 == 0 and i != 0:
es.bulk(bulk_list)
bulk_list = []
i += 1
q_file.close()
es.bulk(bulk_list)
q_file = open('./q_red_with_data.txt')
i = 0
key = None
bulk_list = []
red_index = {"index": {"_index": "i_irelia_state", "_type": "t_red_state"}}
for line in q_file:
import sqlite3
import os
import json
import numpy as np
import datetime
from env.korean_chess import KoreanChess
q_file = open('./q_blue_with_data.txt')
compress_q_file = open('./q_blue_with_data_comp.txt', mode='w')
i = 0
key = None
for line in q_file:
if i % 2 is 0:
key = line.strip()
compress_q_file.write(KoreanChess.compress_state_key(key) + "\n")
else:
compress_q_file.write(line.strip() + "\n")
i += 1
compress_q_file.close()
q_file.close()
q_file = open('./q_red_with_data.txt')
compress_q_file = open('./q_red_with_data_comp.txt', mode='w')
i = 0
key = None
for line in q_file:
if i % 2 is 0:
key = line.strip()
compress_q_file.write(KoreanChess.compress_state_key(key) + "\n")
else:
import numpy as np
import time
import json
from env import korean_chess_util as kcu
from env.korean_chess import KoreanChess
import sys
import sqlite3
conn = sqlite3.connect('q_red.db')
c = conn.cursor()
c.execute("SELECT quality_json FROM t_quality WHERE state_key='r6,r4,r2,r3,1,r3,r4,r2,r6,4,r7,5,r5,5,r5,1,r1,1,r1,1,r1,1,r1,1,r1,18,b1,1,b1,1,b1,1,b1,b1,2,b5,5,b5,5,b7,4,b6,b4,b2,b3,1,b3,b4,b2,b6'")
result = c.fetchone()
print(result)
sys.exit()
ret = KoreanChess.convert_state_map(
'b6,b2,b4,b3,1,b3,b2,b4,b6,4,b7,5,b5,5,b5,1,b1,1,b1,1,b1,1,b1,b1,19,r1,1,r1,1,r1,1,r1,1,r1,1,r5,5,r5,5,r7,4,r6,r4,r2,r3,1,r3,r2,r4,r6')
print(ret)
sys.exit()
Game.register('KoreanChess',
{'position_type': 'random', 'state_list': None,
'init_state': None, 'init_side': 'b'})
env = Game.make('KoreanChess')
ret = env.convert_state_map(
'r6,1,r2,r3,1,r3,r2,1,r6,4,r7,8,r5,r5,r4,6,r4,r1,r1,1,r1,2,r1,14,b6,b1,2,b1,b4,b1,b1,3,b5,b4,1,b5,8,b7,4,b6,1,b2,b3,1,b3,b2,2')
print(ret)
# blue_state = env.reset()