Commit
·
9577796
1
Parent(s):
63fff7e
Delete trainer_state.json
Browse files- trainer_state.json +0 -118
trainer_state.json
DELETED
@@ -1,118 +0,0 @@
|
|
1 |
-
{
|
2 |
-
"best_metric": null,
|
3 |
-
"best_model_checkpoint": null,
|
4 |
-
"epoch": 667.1111111111111,
|
5 |
-
"eval_steps": 500,
|
6 |
-
"global_step": 1501,
|
7 |
-
"is_hyper_param_search": false,
|
8 |
-
"is_local_process_zero": true,
|
9 |
-
"is_world_process_zero": true,
|
10 |
-
"log_history": [
|
11 |
-
{
|
12 |
-
"epoch": 44.44,
|
13 |
-
"learning_rate": 9.333777481678882e-05,
|
14 |
-
"loss": 0.4637,
|
15 |
-
"step": 100
|
16 |
-
},
|
17 |
-
{
|
18 |
-
"epoch": 88.89,
|
19 |
-
"learning_rate": 8.667554963357763e-05,
|
20 |
-
"loss": 0.0349,
|
21 |
-
"step": 200
|
22 |
-
},
|
23 |
-
{
|
24 |
-
"epoch": 133.33,
|
25 |
-
"learning_rate": 8.001332445036642e-05,
|
26 |
-
"loss": 0.034,
|
27 |
-
"step": 300
|
28 |
-
},
|
29 |
-
{
|
30 |
-
"epoch": 177.78,
|
31 |
-
"learning_rate": 7.335109926715523e-05,
|
32 |
-
"loss": 0.0335,
|
33 |
-
"step": 400
|
34 |
-
},
|
35 |
-
{
|
36 |
-
"epoch": 222.22,
|
37 |
-
"learning_rate": 6.668887408394404e-05,
|
38 |
-
"loss": 0.0333,
|
39 |
-
"step": 500
|
40 |
-
},
|
41 |
-
{
|
42 |
-
"epoch": 266.67,
|
43 |
-
"learning_rate": 6.0026648900732854e-05,
|
44 |
-
"loss": 0.0332,
|
45 |
-
"step": 600
|
46 |
-
},
|
47 |
-
{
|
48 |
-
"epoch": 311.11,
|
49 |
-
"learning_rate": 5.336442371752165e-05,
|
50 |
-
"loss": 0.0331,
|
51 |
-
"step": 700
|
52 |
-
},
|
53 |
-
{
|
54 |
-
"epoch": 355.56,
|
55 |
-
"learning_rate": 4.670219853431046e-05,
|
56 |
-
"loss": 0.0328,
|
57 |
-
"step": 800
|
58 |
-
},
|
59 |
-
{
|
60 |
-
"epoch": 400.0,
|
61 |
-
"learning_rate": 4.003997335109927e-05,
|
62 |
-
"loss": 0.0327,
|
63 |
-
"step": 900
|
64 |
-
},
|
65 |
-
{
|
66 |
-
"epoch": 444.44,
|
67 |
-
"learning_rate": 3.337774816788808e-05,
|
68 |
-
"loss": 0.0326,
|
69 |
-
"step": 1000
|
70 |
-
},
|
71 |
-
{
|
72 |
-
"epoch": 488.89,
|
73 |
-
"learning_rate": 2.6715522984676884e-05,
|
74 |
-
"loss": 0.0323,
|
75 |
-
"step": 1100
|
76 |
-
},
|
77 |
-
{
|
78 |
-
"epoch": 533.33,
|
79 |
-
"learning_rate": 2.005329780146569e-05,
|
80 |
-
"loss": 0.0323,
|
81 |
-
"step": 1200
|
82 |
-
},
|
83 |
-
{
|
84 |
-
"epoch": 577.78,
|
85 |
-
"learning_rate": 1.3391072618254496e-05,
|
86 |
-
"loss": 0.0321,
|
87 |
-
"step": 1300
|
88 |
-
},
|
89 |
-
{
|
90 |
-
"epoch": 622.22,
|
91 |
-
"learning_rate": 6.728847435043305e-06,
|
92 |
-
"loss": 0.0321,
|
93 |
-
"step": 1400
|
94 |
-
},
|
95 |
-
{
|
96 |
-
"epoch": 666.67,
|
97 |
-
"learning_rate": 6.662225183211193e-08,
|
98 |
-
"loss": 0.0319,
|
99 |
-
"step": 1500
|
100 |
-
},
|
101 |
-
{
|
102 |
-
"epoch": 667.11,
|
103 |
-
"step": 1501,
|
104 |
-
"total_flos": 4.704291541373952e+16,
|
105 |
-
"train_loss": 0.061614320571872494,
|
106 |
-
"train_runtime": 5482.8502,
|
107 |
-
"train_samples_per_second": 1.095,
|
108 |
-
"train_steps_per_second": 0.274
|
109 |
-
}
|
110 |
-
],
|
111 |
-
"logging_steps": 100,
|
112 |
-
"max_steps": 1501,
|
113 |
-
"num_train_epochs": 751,
|
114 |
-
"save_steps": 500,
|
115 |
-
"total_flos": 4.704291541373952e+16,
|
116 |
-
"trial_name": null,
|
117 |
-
"trial_params": null
|
118 |
-
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|