noflm
commited on
Commit
·
c0564ad
1
Parent(s):
a4f2851
End of training
Browse files
all_results.json
ADDED
@@ -0,0 +1,12 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"epoch": 29.01,
|
3 |
+
"eval_loss": 0.1459457278251648,
|
4 |
+
"eval_runtime": 6.2733,
|
5 |
+
"eval_samples_per_second": 2.869,
|
6 |
+
"eval_steps_per_second": 0.797,
|
7 |
+
"eval_wer": 11.585365853658537,
|
8 |
+
"train_loss": 0.07516751879453659,
|
9 |
+
"train_runtime": 1282.4235,
|
10 |
+
"train_samples_per_second": 6.238,
|
11 |
+
"train_steps_per_second": 0.78
|
12 |
+
}
|
eval_results.json
ADDED
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"epoch": 29.01,
|
3 |
+
"eval_loss": 0.1459457278251648,
|
4 |
+
"eval_runtime": 6.2733,
|
5 |
+
"eval_samples_per_second": 2.869,
|
6 |
+
"eval_steps_per_second": 0.797,
|
7 |
+
"eval_wer": 11.585365853658537
|
8 |
+
}
|
runs/Dec24_14-53-30_DESKTOP-SHN6VHF/events.out.tfevents.1671862581.DESKTOP-SHN6VHF.5414.2
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:c5cdbd8fbd711e34541ce9079ca727db08e831153c4076a867f894d5535cc8c1
|
3 |
+
size 358
|
train_results.json
ADDED
@@ -0,0 +1,7 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"epoch": 29.01,
|
3 |
+
"train_loss": 0.07516751879453659,
|
4 |
+
"train_runtime": 1282.4235,
|
5 |
+
"train_samples_per_second": 6.238,
|
6 |
+
"train_steps_per_second": 0.78
|
7 |
+
}
|
trainer_state.json
ADDED
@@ -0,0 +1,94 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"best_metric": null,
|
3 |
+
"best_model_checkpoint": null,
|
4 |
+
"epoch": 29.014,
|
5 |
+
"global_step": 1000,
|
6 |
+
"is_hyper_param_search": false,
|
7 |
+
"is_local_process_zero": true,
|
8 |
+
"is_world_process_zero": true,
|
9 |
+
"log_history": [
|
10 |
+
{
|
11 |
+
"epoch": 2.03,
|
12 |
+
"learning_rate": 1e-05,
|
13 |
+
"loss": 0.6378,
|
14 |
+
"step": 100
|
15 |
+
},
|
16 |
+
{
|
17 |
+
"epoch": 5.03,
|
18 |
+
"learning_rate": 1e-05,
|
19 |
+
"loss": 0.0803,
|
20 |
+
"step": 200
|
21 |
+
},
|
22 |
+
{
|
23 |
+
"epoch": 8.03,
|
24 |
+
"learning_rate": 1e-05,
|
25 |
+
"loss": 0.0166,
|
26 |
+
"step": 300
|
27 |
+
},
|
28 |
+
{
|
29 |
+
"epoch": 11.03,
|
30 |
+
"learning_rate": 1e-05,
|
31 |
+
"loss": 0.0048,
|
32 |
+
"step": 400
|
33 |
+
},
|
34 |
+
{
|
35 |
+
"epoch": 14.02,
|
36 |
+
"learning_rate": 1e-05,
|
37 |
+
"loss": 0.0037,
|
38 |
+
"step": 500
|
39 |
+
},
|
40 |
+
{
|
41 |
+
"epoch": 17.02,
|
42 |
+
"learning_rate": 1e-05,
|
43 |
+
"loss": 0.0032,
|
44 |
+
"step": 600
|
45 |
+
},
|
46 |
+
{
|
47 |
+
"epoch": 20.02,
|
48 |
+
"learning_rate": 1e-05,
|
49 |
+
"loss": 0.0007,
|
50 |
+
"step": 700
|
51 |
+
},
|
52 |
+
{
|
53 |
+
"epoch": 23.02,
|
54 |
+
"learning_rate": 1e-05,
|
55 |
+
"loss": 0.0011,
|
56 |
+
"step": 800
|
57 |
+
},
|
58 |
+
{
|
59 |
+
"epoch": 26.02,
|
60 |
+
"learning_rate": 1e-05,
|
61 |
+
"loss": 0.0025,
|
62 |
+
"step": 900
|
63 |
+
},
|
64 |
+
{
|
65 |
+
"epoch": 29.01,
|
66 |
+
"learning_rate": 1e-05,
|
67 |
+
"loss": 0.0009,
|
68 |
+
"step": 1000
|
69 |
+
},
|
70 |
+
{
|
71 |
+
"epoch": 29.01,
|
72 |
+
"eval_loss": 0.1459457278251648,
|
73 |
+
"eval_runtime": 5.4122,
|
74 |
+
"eval_samples_per_second": 3.326,
|
75 |
+
"eval_steps_per_second": 0.924,
|
76 |
+
"eval_wer": 11.585365853658537,
|
77 |
+
"step": 1000
|
78 |
+
},
|
79 |
+
{
|
80 |
+
"epoch": 29.01,
|
81 |
+
"step": 1000,
|
82 |
+
"total_flos": 5.1887996928e+17,
|
83 |
+
"train_loss": 0.07516751879453659,
|
84 |
+
"train_runtime": 1282.4235,
|
85 |
+
"train_samples_per_second": 6.238,
|
86 |
+
"train_steps_per_second": 0.78
|
87 |
+
}
|
88 |
+
],
|
89 |
+
"max_steps": 1000,
|
90 |
+
"num_train_epochs": 9223372036854775807,
|
91 |
+
"total_flos": 5.1887996928e+17,
|
92 |
+
"trial_name": null,
|
93 |
+
"trial_params": null
|
94 |
+
}
|