Hugging Face's logo

deepadapt
/
trident-ec

Model card Files
xet
 Community
farguney commited on
Commit
93bf086
·
verified ·
1 Parent(s): 7b7e69d

Upload r4dr_t2p_algo_s0.json with huggingface_hub

Browse files
Files changed (1) hide show
  1. r4dr_t2p_algo_s0.json +125 -0
r4dr_t2p_algo_s0.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
+ 0.001
9
+ ],
10
+ "text_mb": 24.0,
11
+ "d_spine": 384,
12
+ "layers_spine": 8,
13
+ "n_spine": 12018432,
14
+ "tf_mult": 2.0,
15
+ "d_transformer": 624,
16
+ "layers_transformer": 5,
17
+ "n_transformer": 24045216,
18
+ "device": "cuda",
19
+ "thresholds": {
20
+ "pass": 1.05,
21
+ "concern": 1.15
22
+ },
23
+ "corpora": {
24
+ "algo": {
25
+ "spine_lr0.001": {
26
+ "best_bpb": 1.555050253868103,
27
+ "history": [
28
+ {
29
+ "step": 0,
30
+ "train_bits": 8.12976849929531,
31
+ "eval_bpb": 6.19306755065918,
32
+ "best": 6.19306755065918,
33
+ "kBps": 10.5
34
+ },
35
+ {
36
+ "step": 250,
37
+ "train_bits": 2.1256453707365126,
38
+ "eval_bpb": 2.100250720977783,
39
+ "best": 2.100250720977783,
40
+ "kBps": 21.7
41
+ },
42
+ {
43
+ "step": 500,
44
+ "train_bits": 2.0464986108441465,
45
+ "eval_bpb": 2.0480103492736816,
46
+ "best": 2.0480103492736816,
47
+ "kBps": 21.7
48
+ },
49
+ {
50
+ "step": 750,
51
+ "train_bits": 1.6429041723419797,
52
+ "eval_bpb": 1.6335951089859009,
53
+ "best": 1.6335951089859009,
54
+ "kBps": 21.7
55
+ },
56
+ {
57
+ "step": 1000,
58
+ "train_bits": 1.5884018383876504,
59
+ "eval_bpb": 1.57682466506958,
60
+ "best": 1.57682466506958,
61
+ "kBps": 21.7
62
+ },
63
+ {
64
+ "step": 1250,
65
+ "train_bits": 1.559349153142698,
66
+ "eval_bpb": 1.5583819150924683,
67
+ "best": 1.5583819150924683,
68
+ "kBps": 21.7
69
+ },
70
+ {
71
+ "step": 1500,
72
+ "train_bits": 1.557550214614673,
73
+ "eval_bpb": 1.5589181184768677,
74
+ "best": 1.5583819150924683,
75
+ "kBps": 21.7
76
+ },
77
+ {
78
+ "step": 1750,
79
+ "train_bits": 1.5560087341149895,
80
+ "eval_bpb": 1.5574500560760498,
81
+ "best": 1.5574500560760498,
82
+ "kBps": 21.7
83
+ },
84
+ {
85
+ "step": 2000,
86
+ "train_bits": 1.5546904870961262,
87
+ "eval_bpb": 1.55584716796875,
88
+ "best": 1.55584716796875,
89
+ "kBps": 21.7
90
+ },
91
+ {
92
+ "step": 2250,
93
+ "train_bits": 1.557018616240856,
94
+ "eval_bpb": 1.5552558898925781,
95
+ "best": 1.5552558898925781,
96
+ "kBps": 21.7
97
+ },
98
+ {
99
+ "step": 2500,
100
+ "train_bits": 1.5529981778116362,
101
+ "eval_bpb": 1.5552016496658325,
102
+ "best": 1.5552016496658325,
103
+ "kBps": 21.7
104
+ },
105
+ {
106
+ "step": 2750,
107
+ "train_bits": 1.5572064212955983,
108
+ "eval_bpb": 1.5550575256347656,
109
+ "best": 1.5550575256347656,
110
+ "kBps": 21.7
111
+ },
112
+ {
113
+ "step": 2999,
114
+ "train_bits": 1.5532807453070037,
115
+ "eval_bpb": 1.555050253868103,
116
+ "best": 1.555050253868103,
117
+ "kBps": 21.7
118
+ }
119
+ ],
120
+ "wall_s": 2265.3
121
+ }
122
+ }
123
+ },
124
+ "partial": true
125
+ }