-
Notifications
You must be signed in to change notification settings - Fork 2
/
Copy pathhyperparams.py
72 lines (55 loc) · 1.25 KB
/
hyperparams.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
'''
Directory with test levels,
For first submitted network with score 42.66 I used stage3
For second I used stage3 and than stage4
'''
BASE_DIR = '/home/trrrrr/Documents/github/ml/animal_master/rl_animal'
LEARNING_DIR = BASE_DIR + '/configs/learning/competition_configurations/'
'''
Minimum and maximum steps in enviroment
'''
MIN_TIME = 200
MAX_TIME = 1100
'''
Scale of the vel inpuits
'''
VEC_SCALE = [1.0, 1.0, 16.0]
'''
Probability to generate black frames.
Tested every frame for every env. Good value is in [0.0001,0.003]
'''
BLACK_PROBABILITY = 0.0000
'''
When we generating black frames we decidin how many of them to generate.
'''
MIN_BLACK_FRAMES = 3
MAX_BLACK_FRAMES = 9
'''
Will we add penalty for moving back
'''
BACK_MOVE_PENALTY = True
'''
Reward = velocity * BACK_MOVE_COEF if velocity < 0
'''
BACK_MOVE_COEF = 1.0 / 1000.0
'''
Will we add reward for moving up
'''
REWARD_RAMPS = True
'''
Reward = vertical velocity * BACK_MOVE_COEF if velocity > 0
'''
RAMPS_COEF = 1.0 / 100.0
'''
If false then all frames are rgb. If true frist frame is RGB and all other are greyscale
'''
USE_GREYSCALE_OBSES = False
'''
number of visual frames in input
'''
VISUAL_FRAMES_COUNT = 2
'''
number of velocity frames in input
'''
VEL_FRAMES_COUNT = 2
SKIP_FRAMES = 1