xiaoming-leza
commited on
Commit
•
c0e1bc8
1
Parent(s):
e6f04f1
End of training
Browse files
all_results.json
ADDED
@@ -0,0 +1,14 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"epoch": 15.0,
|
3 |
+
"eval_loss": 0.3827837407588959,
|
4 |
+
"eval_runtime": 112.0874,
|
5 |
+
"eval_samples": 1647,
|
6 |
+
"eval_samples_per_second": 14.694,
|
7 |
+
"eval_steps_per_second": 1.838,
|
8 |
+
"eval_wer": 0.34950464712491064,
|
9 |
+
"train_loss": 1.0725414783582774,
|
10 |
+
"train_runtime": 6951.8597,
|
11 |
+
"train_samples": 3478,
|
12 |
+
"train_samples_per_second": 7.504,
|
13 |
+
"train_steps_per_second": 0.235
|
14 |
+
}
|
eval_results.json
ADDED
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"epoch": 15.0,
|
3 |
+
"eval_loss": 0.3827837407588959,
|
4 |
+
"eval_runtime": 112.0874,
|
5 |
+
"eval_samples": 1647,
|
6 |
+
"eval_samples_per_second": 14.694,
|
7 |
+
"eval_steps_per_second": 1.838,
|
8 |
+
"eval_wer": 0.34950464712491064
|
9 |
+
}
|
runs/Jul26_03-55-58_23c4b8f0470c/events.out.tfevents.1690351073.23c4b8f0470c.3028.1
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:247ac971f2d10d2ca8ddf224230886b5ff680cb15637859de2c549d1ec6fa5d8
|
3 |
+
size 406
|
train_results.json
ADDED
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"epoch": 15.0,
|
3 |
+
"train_loss": 1.0725414783582774,
|
4 |
+
"train_runtime": 6951.8597,
|
5 |
+
"train_samples": 3478,
|
6 |
+
"train_samples_per_second": 7.504,
|
7 |
+
"train_steps_per_second": 0.235
|
8 |
+
}
|
trainer_state.json
ADDED
@@ -0,0 +1,187 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"best_metric": null,
|
3 |
+
"best_model_checkpoint": null,
|
4 |
+
"epoch": 15.0,
|
5 |
+
"global_step": 1635,
|
6 |
+
"is_hyper_param_search": false,
|
7 |
+
"is_local_process_zero": true,
|
8 |
+
"is_world_process_zero": true,
|
9 |
+
"log_history": [
|
10 |
+
{
|
11 |
+
"epoch": 0.92,
|
12 |
+
"eval_loss": 3.5826058387756348,
|
13 |
+
"eval_runtime": 112.5008,
|
14 |
+
"eval_samples_per_second": 14.64,
|
15 |
+
"eval_steps_per_second": 1.831,
|
16 |
+
"eval_wer": 1.0,
|
17 |
+
"step": 100
|
18 |
+
},
|
19 |
+
{
|
20 |
+
"epoch": 1.83,
|
21 |
+
"eval_loss": 3.0217819213867188,
|
22 |
+
"eval_runtime": 113.1278,
|
23 |
+
"eval_samples_per_second": 14.559,
|
24 |
+
"eval_steps_per_second": 1.821,
|
25 |
+
"eval_wer": 0.9998978653865795,
|
26 |
+
"step": 200
|
27 |
+
},
|
28 |
+
{
|
29 |
+
"epoch": 2.75,
|
30 |
+
"eval_loss": 0.8984940052032471,
|
31 |
+
"eval_runtime": 111.3066,
|
32 |
+
"eval_samples_per_second": 14.797,
|
33 |
+
"eval_steps_per_second": 1.851,
|
34 |
+
"eval_wer": 0.8035951383924012,
|
35 |
+
"step": 300
|
36 |
+
},
|
37 |
+
{
|
38 |
+
"epoch": 3.67,
|
39 |
+
"eval_loss": 0.5992483496665955,
|
40 |
+
"eval_runtime": 112.2673,
|
41 |
+
"eval_samples_per_second": 14.67,
|
42 |
+
"eval_steps_per_second": 1.835,
|
43 |
+
"eval_wer": 0.619650699622102,
|
44 |
+
"step": 400
|
45 |
+
},
|
46 |
+
{
|
47 |
+
"epoch": 4.59,
|
48 |
+
"learning_rate": 0.00029939999999999996,
|
49 |
+
"loss": 3.1629,
|
50 |
+
"step": 500
|
51 |
+
},
|
52 |
+
{
|
53 |
+
"epoch": 4.59,
|
54 |
+
"eval_loss": 0.49684980511665344,
|
55 |
+
"eval_runtime": 112.6664,
|
56 |
+
"eval_samples_per_second": 14.618,
|
57 |
+
"eval_steps_per_second": 1.828,
|
58 |
+
"eval_wer": 0.5339597589623123,
|
59 |
+
"step": 500
|
60 |
+
},
|
61 |
+
{
|
62 |
+
"epoch": 5.5,
|
63 |
+
"eval_loss": 0.4645964503288269,
|
64 |
+
"eval_runtime": 111.3096,
|
65 |
+
"eval_samples_per_second": 14.797,
|
66 |
+
"eval_steps_per_second": 1.851,
|
67 |
+
"eval_wer": 0.5045449902972117,
|
68 |
+
"step": 600
|
69 |
+
},
|
70 |
+
{
|
71 |
+
"epoch": 6.42,
|
72 |
+
"eval_loss": 0.43159034848213196,
|
73 |
+
"eval_runtime": 113.1768,
|
74 |
+
"eval_samples_per_second": 14.552,
|
75 |
+
"eval_steps_per_second": 1.82,
|
76 |
+
"eval_wer": 0.44254927995097537,
|
77 |
+
"step": 700
|
78 |
+
},
|
79 |
+
{
|
80 |
+
"epoch": 7.34,
|
81 |
+
"eval_loss": 0.44998037815093994,
|
82 |
+
"eval_runtime": 111.8678,
|
83 |
+
"eval_samples_per_second": 14.723,
|
84 |
+
"eval_steps_per_second": 1.841,
|
85 |
+
"eval_wer": 0.4734960678173833,
|
86 |
+
"step": 800
|
87 |
+
},
|
88 |
+
{
|
89 |
+
"epoch": 8.26,
|
90 |
+
"eval_loss": 0.41139060258865356,
|
91 |
+
"eval_runtime": 113.568,
|
92 |
+
"eval_samples_per_second": 14.502,
|
93 |
+
"eval_steps_per_second": 1.814,
|
94 |
+
"eval_wer": 0.41231743437851087,
|
95 |
+
"step": 900
|
96 |
+
},
|
97 |
+
{
|
98 |
+
"epoch": 9.17,
|
99 |
+
"learning_rate": 0.00016810572687224668,
|
100 |
+
"loss": 0.2226,
|
101 |
+
"step": 1000
|
102 |
+
},
|
103 |
+
{
|
104 |
+
"epoch": 9.17,
|
105 |
+
"eval_loss": 0.41622716188430786,
|
106 |
+
"eval_runtime": 112.362,
|
107 |
+
"eval_samples_per_second": 14.658,
|
108 |
+
"eval_steps_per_second": 1.833,
|
109 |
+
"eval_wer": 0.4018997038096211,
|
110 |
+
"step": 1000
|
111 |
+
},
|
112 |
+
{
|
113 |
+
"epoch": 10.09,
|
114 |
+
"eval_loss": 0.3999302089214325,
|
115 |
+
"eval_runtime": 113.8252,
|
116 |
+
"eval_samples_per_second": 14.47,
|
117 |
+
"eval_steps_per_second": 1.81,
|
118 |
+
"eval_wer": 0.38239199264630785,
|
119 |
+
"step": 1100
|
120 |
+
},
|
121 |
+
{
|
122 |
+
"epoch": 11.01,
|
123 |
+
"eval_loss": 0.404796838760376,
|
124 |
+
"eval_runtime": 112.9341,
|
125 |
+
"eval_samples_per_second": 14.584,
|
126 |
+
"eval_steps_per_second": 1.824,
|
127 |
+
"eval_wer": 0.38423041568787664,
|
128 |
+
"step": 1200
|
129 |
+
},
|
130 |
+
{
|
131 |
+
"epoch": 11.93,
|
132 |
+
"eval_loss": 0.3789042532444,
|
133 |
+
"eval_runtime": 112.9251,
|
134 |
+
"eval_samples_per_second": 14.585,
|
135 |
+
"eval_steps_per_second": 1.824,
|
136 |
+
"eval_wer": 0.3602287815340619,
|
137 |
+
"step": 1300
|
138 |
+
},
|
139 |
+
{
|
140 |
+
"epoch": 12.84,
|
141 |
+
"eval_loss": 0.4024040102958679,
|
142 |
+
"eval_runtime": 114.3571,
|
143 |
+
"eval_samples_per_second": 14.402,
|
144 |
+
"eval_steps_per_second": 1.801,
|
145 |
+
"eval_wer": 0.35359003166173014,
|
146 |
+
"step": 1400
|
147 |
+
},
|
148 |
+
{
|
149 |
+
"epoch": 13.76,
|
150 |
+
"learning_rate": 3.594713656387665e-05,
|
151 |
+
"loss": 0.1015,
|
152 |
+
"step": 1500
|
153 |
+
},
|
154 |
+
{
|
155 |
+
"epoch": 13.76,
|
156 |
+
"eval_loss": 0.38988158106803894,
|
157 |
+
"eval_runtime": 113.4392,
|
158 |
+
"eval_samples_per_second": 14.519,
|
159 |
+
"eval_steps_per_second": 1.816,
|
160 |
+
"eval_wer": 0.35747114697170873,
|
161 |
+
"step": 1500
|
162 |
+
},
|
163 |
+
{
|
164 |
+
"epoch": 14.68,
|
165 |
+
"eval_loss": 0.3801647126674652,
|
166 |
+
"eval_runtime": 112.3523,
|
167 |
+
"eval_samples_per_second": 14.659,
|
168 |
+
"eval_steps_per_second": 1.834,
|
169 |
+
"eval_wer": 0.3489939740578082,
|
170 |
+
"step": 1600
|
171 |
+
},
|
172 |
+
{
|
173 |
+
"epoch": 15.0,
|
174 |
+
"step": 1635,
|
175 |
+
"total_flos": 6.468063251673315e+18,
|
176 |
+
"train_loss": 1.0725414783582774,
|
177 |
+
"train_runtime": 6951.8597,
|
178 |
+
"train_samples_per_second": 7.504,
|
179 |
+
"train_steps_per_second": 0.235
|
180 |
+
}
|
181 |
+
],
|
182 |
+
"max_steps": 1635,
|
183 |
+
"num_train_epochs": 15,
|
184 |
+
"total_flos": 6.468063251673315e+18,
|
185 |
+
"trial_name": null,
|
186 |
+
"trial_params": null
|
187 |
+
}
|