-
Notifications
You must be signed in to change notification settings - Fork 0
/
utils.py
73 lines (58 loc) · 2.59 KB
/
utils.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
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
import numpy as np
class RunningMeanStd(object):
# https://en.wikipedia.org/wiki/Algorithms_for_calculating_variance#Parallel_algorithm
def __init__(self, epsilon=0, shape=()):
self.mean = np.zeros(shape, 'float64')
self.var = np.zeros(shape, 'float64') # <-- np.ones
self.count = epsilon
def update(self, x):
batch_count = x.shape[0]
batch_mean = np.mean(x, axis=0)
ddof = 0 if batch_count == 1 else 1
batch_var = np.var(x, axis=0, ddof=ddof)
self.update_from_moments(batch_mean, batch_var, batch_count)
def update_from_moments(self, batch_mean, batch_var, batch_count):
delta = batch_mean - self.mean
tot_count = self.count + batch_count
new_mean = self.mean + delta * batch_count / tot_count
m_a = self.var * (self.count - 1)
m_b = batch_var * (batch_count - 1)
M2 = m_a + m_b + np.square(delta) * self.count * batch_count / (self.count + batch_count)
new_var = M2 / (self.count + batch_count - 1)
new_count = batch_count + self.count
self.mean = new_mean
self.var = new_var
self.count = new_count
class RewardForwardFilter(object):
def __init__(self, gamma):
self.rewems = None
self.gamma = gamma
def update(self, rews):
if self.rewems is None:
self.rewems = rews
else:
self.rewems = self.rewems * self.gamma + rews
return self.rewems
class RunningMeanStd_openAI(object):
# https://en.wikipedia.org/wiki/Algorithms_for_calculating_variance#Parallel_algorithm
def __init__(self, epsilon=1e-4, shape=()):
self.mean = np.zeros(shape, 'float64')
self.var = np.ones(shape, 'float64')
self.count = epsilon
def update(self, x):
batch_mean = np.mean(x, axis=0)
batch_var = np.var(x, axis=0)
batch_count = x.shape[0]
self.update_from_moments(batch_mean, batch_var, batch_count)
def update_from_moments(self, batch_mean, batch_var, batch_count):
delta = batch_mean - self.mean
tot_count = self.count + batch_count
new_mean = self.mean + delta * batch_count / tot_count
m_a = self.var * (self.count)
m_b = batch_var * (batch_count)
M2 = m_a + m_b + np.square(delta) * self.count * batch_count / (self.count + batch_count)
new_var = M2 / (self.count + batch_count)
new_count = batch_count + self.count
self.mean = new_mean
self.var = new_var
self.count = new_count