# -*- coding: utf-8 -*-
import time
import threading
import redis
from ..libs import utils, replay_memory
class Replay(threading.Thread):
"""Replay of Ape-X
Args:
size (int, optional): size of memory
connect (redis.StrictRedis, optional): Redis client object
use_compress (bool, optional): use the compressed memory for saved memory
"""
def __init__(self, size=50000, connect=redis.StrictRedis(host='localhost'),
use_compress=False):
super(Replay, self).__init__()
self.setDaemon(True)
self._memory = replay_memory.PrioritizedMemory(size, use_compress)
self._connect = connect
self._connect.delete('experience')
self._lock = threading.Lock()
def run(self):
while True:
pipe = self._connect.pipeline()
pipe.lrange('experience', 0, -1)
pipe.ltrim('experience', -1, 0)
data = pipe.execute()[0]
if not data is None:
for d in data:
t, p = utils.loads(d)
with self._lock:
self._memory.push(t, p)
time.sleep(0.01)
def update_priorities(self, indices, priorities):
with self._lock:
self._memory.update_priorities(indices, priorities)
def remove_to_fit(self):
with self._lock:
self._memory.remove_to_fit()
def sample(self, batch_size):
with self._lock:
return self._memory.sample(batch_size)
def __len__(self):
return len(self._memory)
@property
def total_prios(self):
return self._memory.total_prios()
