farguney commited on
Commit
3af0b4c
·
verified ·
1 Parent(s): 34c30cf

Upload sp2_spine_text_s0_lr2.json with huggingface_hub

Browse files
Files changed (1) hide show
  1. sp2_spine_text_s0_lr2.json +125 -0
sp2_spine_text_s0_lr2.json ADDED
@@ -0,0 +1,125 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "screen": "SP-2-SPINE-LM",
3
+ "seed": 0,
4
+ "steps": 3000,
5
+ "batch": 32,
6
+ "block": 512,
7
+ "lrs": [
8
+ 0.0003
9
+ ],
10
+ "text_mb": 24.0,
11
+ "d_spine": 384,
12
+ "layers_spine": 8,
13
+ "n_spine": 12018432,
14
+ "tf_mult": 1.0,
15
+ "d_transformer": 400,
16
+ "layers_transformer": 6,
17
+ "n_transformer": 11962800,
18
+ "device": "cuda",
19
+ "thresholds": {
20
+ "pass": 1.05,
21
+ "concern": 1.15
22
+ },
23
+ "corpora": {
24
+ "text": {
25
+ "spine_lr0.0003": {
26
+ "best_bpb": 1.6920182704925537,
27
+ "history": [
28
+ {
29
+ "step": 0,
30
+ "train_bits": 8.10444164619865,
31
+ "eval_bpb": 6.368494033813477,
32
+ "best": 6.368494033813477,
33
+ "kBps": 10.9
34
+ },
35
+ {
36
+ "step": 250,
37
+ "train_bits": 2.0672142651058736,
38
+ "eval_bpb": 2.1344122886657715,
39
+ "best": 2.1344122886657715,
40
+ "kBps": 21.1
41
+ },
42
+ {
43
+ "step": 500,
44
+ "train_bits": 1.9160518339565165,
45
+ "eval_bpb": 1.9972063302993774,
46
+ "best": 1.9972063302993774,
47
+ "kBps": 21.1
48
+ },
49
+ {
50
+ "step": 750,
51
+ "train_bits": 1.7798878378062541,
52
+ "eval_bpb": 1.9081671237945557,
53
+ "best": 1.9081671237945557,
54
+ "kBps": 21.1
55
+ },
56
+ {
57
+ "step": 1000,
58
+ "train_bits": 1.7443163221629978,
59
+ "eval_bpb": 1.852670669555664,
60
+ "best": 1.852670669555664,
61
+ "kBps": 21.1
62
+ },
63
+ {
64
+ "step": 1250,
65
+ "train_bits": 1.6570211962554025,
66
+ "eval_bpb": 1.8103423118591309,
67
+ "best": 1.8103423118591309,
68
+ "kBps": 21.1
69
+ },
70
+ {
71
+ "step": 1500,
72
+ "train_bits": 1.6759471990522512,
73
+ "eval_bpb": 1.7894508838653564,
74
+ "best": 1.7894508838653564,
75
+ "kBps": 21.1
76
+ },
77
+ {
78
+ "step": 1750,
79
+ "train_bits": 1.6142092110108217,
80
+ "eval_bpb": 1.745488166809082,
81
+ "best": 1.745488166809082,
82
+ "kBps": 21.1
83
+ },
84
+ {
85
+ "step": 2000,
86
+ "train_bits": 1.6002025999592528,
87
+ "eval_bpb": 1.721181035041809,
88
+ "best": 1.721181035041809,
89
+ "kBps": 21.1
90
+ },
91
+ {
92
+ "step": 2250,
93
+ "train_bits": 1.5390813417038176,
94
+ "eval_bpb": 1.7037999629974365,
95
+ "best": 1.7037999629974365,
96
+ "kBps": 21.1
97
+ },
98
+ {
99
+ "step": 2500,
100
+ "train_bits": 1.4838240039122006,
101
+ "eval_bpb": 1.6920182704925537,
102
+ "best": 1.6920182704925537,
103
+ "kBps": 21.1
104
+ },
105
+ {
106
+ "step": 2750,
107
+ "train_bits": 1.5799485471324417,
108
+ "eval_bpb": 1.6952470541000366,
109
+ "best": 1.6920182704925537,
110
+ "kBps": 21.1
111
+ },
112
+ {
113
+ "step": 2999,
114
+ "train_bits": 1.4944028287494229,
115
+ "eval_bpb": 1.6929237842559814,
116
+ "best": 1.6920182704925537,
117
+ "kBps": 21.1
118
+ }
119
+ ],
120
+ "wall_s": 2334.0
121
+ }
122
+ }
123
+ },
124
+ "partial": true
125
+ }