Update log_history.json
Browse files- log_history.json +149 -0
log_history.json
ADDED
@@ -0,0 +1,149 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
[
|
2 |
+
{
|
3 |
+
"loss": 0.712146484375,
|
4 |
+
"learning_rate": 9.526784024228659e-05,
|
5 |
+
"epoch": 0.1419547164454539,
|
6 |
+
"total_flos": 3042241413120000,
|
7 |
+
"step": 500
|
8 |
+
},
|
9 |
+
{
|
10 |
+
"loss": 0.421217041015625,
|
11 |
+
"learning_rate": 9.053568048457316e-05,
|
12 |
+
"epoch": 0.2839094328909078,
|
13 |
+
"total_flos": 6084482826240000,
|
14 |
+
"step": 1000
|
15 |
+
},
|
16 |
+
{
|
17 |
+
"loss": 0.39531640625,
|
18 |
+
"learning_rate": 8.580352072685975e-05,
|
19 |
+
"epoch": 0.4258641493363617,
|
20 |
+
"total_flos": 9126724239360000,
|
21 |
+
"step": 1500
|
22 |
+
},
|
23 |
+
{
|
24 |
+
"loss": 0.3591341552734375,
|
25 |
+
"learning_rate": 8.107136096914632e-05,
|
26 |
+
"epoch": 0.5678188657818156,
|
27 |
+
"total_flos": 12168965652480000,
|
28 |
+
"step": 2000
|
29 |
+
},
|
30 |
+
{
|
31 |
+
"loss": 0.3376422119140625,
|
32 |
+
"learning_rate": 7.63392012114329e-05,
|
33 |
+
"epoch": 0.7097735822272695,
|
34 |
+
"total_flos": 15211207065600000,
|
35 |
+
"step": 2500
|
36 |
+
},
|
37 |
+
{
|
38 |
+
"loss": 0.328197998046875,
|
39 |
+
"learning_rate": 7.160704145371948e-05,
|
40 |
+
"epoch": 0.8517282986727234,
|
41 |
+
"total_flos": 18253448478720000,
|
42 |
+
"step": 3000
|
43 |
+
},
|
44 |
+
{
|
45 |
+
"loss": 0.3190537109375,
|
46 |
+
"learning_rate": 6.687488169600607e-05,
|
47 |
+
"epoch": 0.9936830151181772,
|
48 |
+
"total_flos": 21295689891840000,
|
49 |
+
"step": 3500
|
50 |
+
},
|
51 |
+
{
|
52 |
+
"loss": 0.28114892578125,
|
53 |
+
"learning_rate": 6.214272193829264e-05,
|
54 |
+
"epoch": 1.1357087089218538,
|
55 |
+
"total_flos": 24339072145489920,
|
56 |
+
"step": 4000
|
57 |
+
},
|
58 |
+
{
|
59 |
+
"loss": 0.2707734375,
|
60 |
+
"learning_rate": 5.7410562180579217e-05,
|
61 |
+
"epoch": 1.2776634253673078,
|
62 |
+
"total_flos": 27381313558609920,
|
63 |
+
"step": 4500
|
64 |
+
},
|
65 |
+
{
|
66 |
+
"loss": 0.26517333984375,
|
67 |
+
"learning_rate": 5.26784024228658e-05,
|
68 |
+
"epoch": 1.4196181418127618,
|
69 |
+
"total_flos": 30423554971729920,
|
70 |
+
"step": 5000
|
71 |
+
},
|
72 |
+
{
|
73 |
+
"loss": 0.262806884765625,
|
74 |
+
"learning_rate": 4.794624266515238e-05,
|
75 |
+
"epoch": 1.5615728582582156,
|
76 |
+
"total_flos": 33465796384849920,
|
77 |
+
"step": 5500
|
78 |
+
},
|
79 |
+
{
|
80 |
+
"loss": 0.255599853515625,
|
81 |
+
"learning_rate": 4.321408290743896e-05,
|
82 |
+
"epoch": 1.7035275747036696,
|
83 |
+
"total_flos": 36508037797969920,
|
84 |
+
"step": 6000
|
85 |
+
},
|
86 |
+
{
|
87 |
+
"loss": 0.2491875,
|
88 |
+
"learning_rate": 3.848192314972553e-05,
|
89 |
+
"epoch": 1.8454822911491235,
|
90 |
+
"total_flos": 39550279211089920,
|
91 |
+
"step": 6500
|
92 |
+
},
|
93 |
+
{
|
94 |
+
"loss": 0.25085498046875,
|
95 |
+
"learning_rate": 3.374976339201212e-05,
|
96 |
+
"epoch": 1.9874370075945773,
|
97 |
+
"total_flos": 42592520624209920,
|
98 |
+
"step": 7000
|
99 |
+
},
|
100 |
+
{
|
101 |
+
"loss": 0.22507177734375,
|
102 |
+
"learning_rate": 2.9017603634298694e-05,
|
103 |
+
"epoch": 2.129462701398254,
|
104 |
+
"total_flos": 45635902877859840,
|
105 |
+
"step": 7500
|
106 |
+
},
|
107 |
+
{
|
108 |
+
"loss": 0.2233544921875,
|
109 |
+
"learning_rate": 2.4285443876585277e-05,
|
110 |
+
"epoch": 2.2714174178437077,
|
111 |
+
"total_flos": 48678144290979840,
|
112 |
+
"step": 8000
|
113 |
+
},
|
114 |
+
{
|
115 |
+
"loss": 0.21958740234375,
|
116 |
+
"learning_rate": 1.9553284118871853e-05,
|
117 |
+
"epoch": 2.413372134289162,
|
118 |
+
"total_flos": 51720385704099840,
|
119 |
+
"step": 8500
|
120 |
+
},
|
121 |
+
{
|
122 |
+
"loss": 0.22100439453125,
|
123 |
+
"learning_rate": 1.4821124361158433e-05,
|
124 |
+
"epoch": 2.5553268507346156,
|
125 |
+
"total_flos": 54762627117219840,
|
126 |
+
"step": 9000
|
127 |
+
},
|
128 |
+
{
|
129 |
+
"loss": 0.21701953125,
|
130 |
+
"learning_rate": 1.0088964603445012e-05,
|
131 |
+
"epoch": 2.6972815671800694,
|
132 |
+
"total_flos": 57804868530339840,
|
133 |
+
"step": 9500
|
134 |
+
},
|
135 |
+
{
|
136 |
+
"loss": 0.21920361328125,
|
137 |
+
"learning_rate": 5.356804845731592e-06,
|
138 |
+
"epoch": 2.8392362836255236,
|
139 |
+
"total_flos": 60847109943459840,
|
140 |
+
"step": 10000
|
141 |
+
},
|
142 |
+
{
|
143 |
+
"loss": 0.21878466796875,
|
144 |
+
"learning_rate": 6.246450880181715e-07,
|
145 |
+
"epoch": 2.9811910000709774,
|
146 |
+
"total_flos": 63889351356579840,
|
147 |
+
"step": 10500
|
148 |
+
}
|
149 |
+
]
|