-
Notifications
You must be signed in to change notification settings - Fork 9
/
Copy pathmodel_test_log
86 lines (76 loc) · 1023 Bytes
/
model_test_log
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
74
75
76
77
78
79
80
81
82
83
84
85
86
# feature scaling test
## 200 epoch
- 0.1~0.7
695867
696315
696075
696223
697029
- 1~7 / 7
706036
701323
70208
70278
702206
- 0.4~1.0
695979
695641
696154
696012
69607
- 0.143,0.286...1.0
697671
69853
696594
69821
697419
- 0.70 ~ 1.0
69548
695531
695265
696716
695664
- 0.94 ~ 1.0
694733
695198
695045
694799
695368
- 0.97 ~ 1.0
695183
694621
695185
695387
694911
- 0.994 ~ 1.0
694767
695181
695593
694912
694932
# use or notuse bias
- 0.4 ~ 1.0
695435
696293
696253
696684
695328
# test list
feature scaling test -> effect middle, more good
use or notuse bias -> no effect
bm -> 0.7
matrix output vs vector output -> vector is more more better
value and policy network test -> similar~ more bad little
last conv relu -> no effect
reshape fc num_classes *2 -> effect little, more good
no fc -> effect big, more shit
fc 2layers -> effect litte, more bat
regulizer -> similar , no overfit
many layers
resnet layers
many state
many values
value decay
with color channel
with real one game data
variational batch size