def __init__(self, buffer_size):
self.buffer_size = buffer_size
self.num_experiences = 0
#self.buffer = deque()
conf = {
'size': 10000,
'learn_start': 32,
'partition_num': 32,
'total_step': 10000,
'batch_size': 32
}
self.replay_memory = rank_based.Experience(conf)
def t():
conf = {
'size': 50,
'learn_start': 10,
'partition_num': 5,
'total_step': 100,
'batch_size': 4
}
experience = rank_based.Experience(conf)
# insert to experience
print('test insert experience')
for i in range(1, 51):
# tuple, like(state_t, a, r, state_t_1, t)
to_insert = (i, 1, 1, i, 1)
experience.store(to_insert)
print(experience.priority_queue)
print(experience._experience[1])
print(experience._experience[2])
print('test replace')
to_insert = (51, 1, 1, 51, 1)
experience.store(to_insert)
print(experience.priority_queue)
print(experience._experience[1])
print(experience._experience[2])
# sample
print('test sample')
sample, w, e_id = experience.sample(51)
print(sample)
print(w)
print(e_id)
# update delta to priority
print('test update delta')
delta = [v for v in range(1, 5)]
experience.update_priority(e_id, delta)
print(experience.priority_queue)
sample, w, e_id = experience.sample(51)
print(sample)
print(w)
print(e_id)
# rebalance
print('test rebalance')
experience.rebalance()
print(experience.priority_queue)