/
basic.py
47 lines (34 loc) · 1.27 KB
/
basic.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
import random
import numpy as np
import tensorflow as tf
from .registry import register
from ....utils.utils import ModeKeys
@register
def random_action(hparams, distribution, worker_id=0):
return random.randint(0, hparams.num_actions - 1)
@register
def epsilon_action(hparams, distribution, worker_id=0):
# perform random action during training only
if (hparams.mode[worker_id] == ModeKeys.TRAIN and
random.random() < hparams.epsilon[worker_id]):
return random_action(hparams, distribution)
else:
return max_action(hparams, distribution)
@register
def max_action(hparams, distribution, worker_id=0):
return np.argmax(distribution)
@register
def non_uniform_random_action(hparams, distribution, worker_id=0):
return np.random.choice(range(hparams.num_actions), p=distribution.ravel())
@register
def uniform_random_action(hparams, distribution, worker_id=0):
if hparams.mode[worker_id] == ModeKeys.TRAIN:
h = np.random.uniform(size=distribution.shape)
return np.argmax(distribution - np.log(-np.log(h)))
else:
return max_action(hparams, distribution, worker_id=0)
@register
def normal_noise_action(hparams, action, worker_id=0):
return np.clip(
np.random.normal(action, hparams.variance), hparams.action_low,
hparams.action_high)