Hugging Face's logo

AbstractPhil
/
alephlm-0

Feature Extraction
TensorBoard
English
sentence-similarity
consensus-distillation
mixture-of-experts
sparse-routing
geometric-deep-learning
amoe
research-log
Model card Files
xet
 Metrics Community
AbstractPhil commited on
Commit
b76c0a7
·
verified ·
1 Parent(s): 21857ce

A2 toggle ladder: alephlm0-a3_random-s0 (6 snapshots)

Browse files
Files changed (1) hide show
  1. analysis/toggle_ladder_alephlm0-a3_random-s0.json +163 -0
analysis/toggle_ladder_alephlm0-a3_random-s0.json ADDED
@@ -0,0 +1,163 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "run": "alephlm0-a3_random-s0",
3
+ "arm": "a3_random",
4
+ "rows": [
5
+ {
6
+ "step": 1055,
7
+ "on": {
8
+ "STS-B": 0.3803,
9
+ "SICK-R": 0.5639,
10
+ "STS12": 0.3849,
11
+ "STS13": 0.3866,
12
+ "STS14": 0.3805,
13
+ "STS15": 0.5463,
14
+ "STS16": 0.3898,
15
+ "BIOSSES": 0.4754
16
+ },
17
+ "off": {
18
+ "STS-B": 0.3718,
19
+ "SICK-R": 0.5612,
20
+ "STS12": 0.375,
21
+ "STS13": 0.3728,
22
+ "STS14": 0.3727,
23
+ "STS15": 0.5397,
24
+ "STS16": 0.3912,
25
+ "BIOSSES": 0.434
26
+ },
27
+ "mean_on": 0.4385,
28
+ "mean_off": 0.4273,
29
+ "toggle": 0.0112
30
+ },
31
+ {
32
+ "step": 5436,
33
+ "on": {
34
+ "STS-B": 0.5364,
35
+ "SICK-R": 0.6413,
36
+ "STS12": 0.4793,
37
+ "STS13": 0.4891,
38
+ "STS14": 0.5009,
39
+ "STS15": 0.6879,
40
+ "STS16": 0.5732,
41
+ "BIOSSES": 0.4843
42
+ },
43
+ "off": {
44
+ "STS-B": 0.5096,
45
+ "SICK-R": 0.6132,
46
+ "STS12": 0.439,
47
+ "STS13": 0.4906,
48
+ "STS14": 0.5006,
49
+ "STS15": 0.6684,
50
+ "STS16": 0.5511,
51
+ "BIOSSES": 0.4687
52
+ },
53
+ "mean_on": 0.5491,
54
+ "mean_off": 0.5301,
55
+ "toggle": 0.0189
56
+ },
57
+ {
58
+ "step": 8691,
59
+ "on": {
60
+ "STS-B": 0.5661,
61
+ "SICK-R": 0.6607,
62
+ "STS12": 0.5029,
63
+ "STS13": 0.5618,
64
+ "STS14": 0.5361,
65
+ "STS15": 0.699,
66
+ "STS16": 0.6235,
67
+ "BIOSSES": 0.5313
68
+ },
69
+ "off": {
70
+ "STS-B": 0.5292,
71
+ "SICK-R": 0.6211,
72
+ "STS12": 0.4548,
73
+ "STS13": 0.5374,
74
+ "STS14": 0.5378,
75
+ "STS15": 0.6747,
76
+ "STS16": 0.5926,
77
+ "BIOSSES": 0.4977
78
+ },
79
+ "mean_on": 0.5852,
80
+ "mean_off": 0.5557,
81
+ "toggle": 0.0295
82
+ },
83
+ {
84
+ "step": 17404,
85
+ "on": {
86
+ "STS-B": 0.5719,
87
+ "SICK-R": 0.6558,
88
+ "STS12": 0.5124,
89
+ "STS13": 0.5996,
90
+ "STS14": 0.5439,
91
+ "STS15": 0.7105,
92
+ "STS16": 0.6657,
93
+ "BIOSSES": 0.563
94
+ },
95
+ "off": {
96
+ "STS-B": 0.5679,
97
+ "SICK-R": 0.6193,
98
+ "STS12": 0.4776,
99
+ "STS13": 0.5637,
100
+ "STS14": 0.5506,
101
+ "STS15": 0.6827,
102
+ "STS16": 0.6404,
103
+ "BIOSSES": 0.4183
104
+ },
105
+ "mean_on": 0.6029,
106
+ "mean_off": 0.5651,
107
+ "toggle": 0.0378
108
+ },
109
+ {
110
+ "step": 30552,
111
+ "on": {
112
+ "STS-B": 0.5771,
113
+ "SICK-R": 0.6572,
114
+ "STS12": 0.4972,
115
+ "STS13": 0.5957,
116
+ "STS14": 0.5499,
117
+ "STS15": 0.7131,
118
+ "STS16": 0.6758,
119
+ "BIOSSES": 0.5489
120
+ },
121
+ "off": {
122
+ "STS-B": 0.5891,
123
+ "SICK-R": 0.6278,
124
+ "STS12": 0.4869,
125
+ "STS13": 0.5799,
126
+ "STS14": 0.5532,
127
+ "STS15": 0.6907,
128
+ "STS16": 0.6359,
129
+ "BIOSSES": 0.4589
130
+ },
131
+ "mean_on": 0.6019,
132
+ "mean_off": 0.5778,
133
+ "toggle": 0.0241
134
+ },
135
+ {
136
+ "step": 61350,
137
+ "on": {
138
+ "STS-B": 0.5704,
139
+ "SICK-R": 0.6539,
140
+ "STS12": 0.4962,
141
+ "STS13": 0.6,
142
+ "STS14": 0.5447,
143
+ "STS15": 0.7102,
144
+ "STS16": 0.6804,
145
+ "BIOSSES": 0.5698
146
+ },
147
+ "off": {
148
+ "STS-B": 0.5764,
149
+ "SICK-R": 0.631,
150
+ "STS12": 0.4947,
151
+ "STS13": 0.5734,
152
+ "STS14": 0.5486,
153
+ "STS15": 0.6853,
154
+ "STS16": 0.6305,
155
+ "BIOSSES": 0.476
156
+ },
157
+ "mean_on": 0.6032,
158
+ "mean_off": 0.577,
159
+ "toggle": 0.0262
160
+ }
161
+ ],
162
+ "hub_repo": "AbstractPhil/alephlm-0"
163
+ }