ivangtorre
commited on
Commit
•
ce8b514
1
Parent(s):
7b77730
Add model files
Browse files- pytorch_model.bin +1 -1
- trainer_state.json +105 -42
- training_args.bin +1 -1
- vocab.json +1 -1
pytorch_model.bin
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 1262055847
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:a7d28736df2487e8090bdd671f4677a7595b91eb816aaaf41581b6e02397000e
|
3 |
size 1262055847
|
trainer_state.json
CHANGED
@@ -1,79 +1,142 @@
|
|
1 |
{
|
2 |
"best_metric": null,
|
3 |
"best_model_checkpoint": null,
|
4 |
-
"epoch":
|
5 |
-
"global_step":
|
6 |
"is_hyper_param_search": false,
|
7 |
"is_local_process_zero": true,
|
8 |
"is_world_process_zero": true,
|
9 |
"log_history": [
|
10 |
{
|
11 |
-
"epoch":
|
12 |
-
"learning_rate": 0.
|
13 |
-
"loss": 1.
|
14 |
"step": 500
|
15 |
},
|
16 |
{
|
17 |
-
"epoch":
|
18 |
-
"eval_loss": 0.
|
19 |
-
"eval_runtime":
|
20 |
-
"eval_samples_per_second":
|
21 |
-
"eval_wer": 0.
|
22 |
"step": 500
|
23 |
},
|
24 |
{
|
25 |
-
"epoch":
|
26 |
-
"learning_rate": 0.
|
27 |
-
"loss": 0.
|
28 |
"step": 1000
|
29 |
},
|
30 |
{
|
31 |
-
"epoch":
|
32 |
-
"eval_loss": 0.
|
33 |
-
"eval_runtime":
|
34 |
-
"eval_samples_per_second":
|
35 |
-
"eval_wer": 0.
|
36 |
"step": 1000
|
37 |
},
|
38 |
{
|
39 |
-
"epoch":
|
40 |
-
"learning_rate":
|
41 |
-
"loss": 0.
|
42 |
"step": 1500
|
43 |
},
|
44 |
{
|
45 |
-
"epoch":
|
46 |
-
"eval_loss": 0.
|
47 |
-
"eval_runtime":
|
48 |
-
"eval_samples_per_second": 6.
|
49 |
-
"eval_wer": 0.
|
50 |
"step": 1500
|
51 |
},
|
52 |
{
|
53 |
-
"epoch":
|
54 |
-
"learning_rate":
|
55 |
-
"loss": 0.
|
56 |
"step": 2000
|
57 |
},
|
58 |
{
|
59 |
-
"epoch":
|
60 |
-
"eval_loss": 0.
|
61 |
-
"eval_runtime":
|
62 |
-
"eval_samples_per_second": 7.
|
63 |
-
"eval_wer": 0.
|
64 |
"step": 2000
|
65 |
},
|
66 |
{
|
67 |
-
"epoch":
|
68 |
-
"
|
69 |
-
"
|
70 |
-
"
|
71 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
72 |
}
|
73 |
],
|
74 |
-
"max_steps":
|
75 |
-
"num_train_epochs":
|
76 |
-
"total_flos":
|
77 |
"trial_name": null,
|
78 |
"trial_params": null
|
79 |
}
|
1 |
{
|
2 |
"best_metric": null,
|
3 |
"best_model_checkpoint": null,
|
4 |
+
"epoch": 34.09068010075567,
|
5 |
+
"global_step": 4500,
|
6 |
"is_hyper_param_search": false,
|
7 |
"is_local_process_zero": true,
|
8 |
"is_world_process_zero": true,
|
9 |
"log_history": [
|
10 |
{
|
11 |
+
"epoch": 3.79,
|
12 |
+
"learning_rate": 0.00039436231366470835,
|
13 |
+
"loss": 1.5121,
|
14 |
"step": 500
|
15 |
},
|
16 |
{
|
17 |
+
"epoch": 3.79,
|
18 |
+
"eval_loss": 0.186998650431633,
|
19 |
+
"eval_runtime": 714.138,
|
20 |
+
"eval_samples_per_second": 7.242,
|
21 |
+
"eval_wer": 0.39502654080485317,
|
22 |
"step": 500
|
23 |
},
|
24 |
{
|
25 |
+
"epoch": 7.57,
|
26 |
+
"learning_rate": 0.00037776708973098473,
|
27 |
+
"loss": 0.1108,
|
28 |
"step": 1000
|
29 |
},
|
30 |
{
|
31 |
+
"epoch": 7.57,
|
32 |
+
"eval_loss": 0.16734005510807037,
|
33 |
+
"eval_runtime": 670.8576,
|
34 |
+
"eval_samples_per_second": 7.71,
|
35 |
+
"eval_wer": 0.33428339826896425,
|
36 |
"step": 1000
|
37 |
},
|
38 |
{
|
39 |
+
"epoch": 11.36,
|
40 |
+
"learning_rate": 0.0003511499148708517,
|
41 |
+
"loss": 0.0666,
|
42 |
"step": 1500
|
43 |
},
|
44 |
{
|
45 |
+
"epoch": 11.36,
|
46 |
+
"eval_loss": 0.17631034553050995,
|
47 |
+
"eval_runtime": 758.94,
|
48 |
+
"eval_samples_per_second": 6.815,
|
49 |
+
"eval_wer": 0.3243469393091546,
|
50 |
"step": 1500
|
51 |
},
|
52 |
{
|
53 |
+
"epoch": 15.15,
|
54 |
+
"learning_rate": 0.00031601138191423965,
|
55 |
+
"loss": 0.0459,
|
56 |
"step": 2000
|
57 |
},
|
58 |
{
|
59 |
+
"epoch": 15.15,
|
60 |
+
"eval_loss": 0.2086898684501648,
|
61 |
+
"eval_runtime": 682.0381,
|
62 |
+
"eval_samples_per_second": 7.583,
|
63 |
+
"eval_wer": 0.32364092775148395,
|
64 |
"step": 2000
|
65 |
},
|
66 |
{
|
67 |
+
"epoch": 18.94,
|
68 |
+
"learning_rate": 0.0002743324911320655,
|
69 |
+
"loss": 0.0358,
|
70 |
+
"step": 2500
|
71 |
+
},
|
72 |
+
{
|
73 |
+
"epoch": 18.94,
|
74 |
+
"eval_loss": 0.1979219615459442,
|
75 |
+
"eval_runtime": 700.5817,
|
76 |
+
"eval_samples_per_second": 7.382,
|
77 |
+
"eval_wer": 0.3092069136835499,
|
78 |
+
"step": 2500
|
79 |
+
},
|
80 |
+
{
|
81 |
+
"epoch": 22.73,
|
82 |
+
"learning_rate": 0.00022846296765465705,
|
83 |
+
"loss": 0.0285,
|
84 |
+
"step": 3000
|
85 |
+
},
|
86 |
+
{
|
87 |
+
"epoch": 22.73,
|
88 |
+
"eval_loss": 0.19415870308876038,
|
89 |
+
"eval_runtime": 1037.6087,
|
90 |
+
"eval_samples_per_second": 4.985,
|
91 |
+
"eval_wer": 0.3047093585754256,
|
92 |
+
"step": 3000
|
93 |
+
},
|
94 |
+
{
|
95 |
+
"epoch": 26.51,
|
96 |
+
"learning_rate": 0.0001809887913391635,
|
97 |
+
"loss": 0.0237,
|
98 |
+
"step": 3500
|
99 |
+
},
|
100 |
+
{
|
101 |
+
"epoch": 26.51,
|
102 |
+
"eval_loss": 0.1929541826248169,
|
103 |
+
"eval_runtime": 764.504,
|
104 |
+
"eval_samples_per_second": 6.765,
|
105 |
+
"eval_wer": 0.30026410061972125,
|
106 |
+
"step": 3500
|
107 |
+
},
|
108 |
+
{
|
109 |
+
"epoch": 30.3,
|
110 |
+
"learning_rate": 0.0001345864073365157,
|
111 |
+
"loss": 0.0205,
|
112 |
+
"step": 4000
|
113 |
+
},
|
114 |
+
{
|
115 |
+
"epoch": 30.3,
|
116 |
+
"eval_loss": 0.2030186504125595,
|
117 |
+
"eval_runtime": 1540.1942,
|
118 |
+
"eval_samples_per_second": 3.358,
|
119 |
+
"eval_wer": 0.29338702507648456,
|
120 |
+
"step": 4000
|
121 |
+
},
|
122 |
+
{
|
123 |
+
"epoch": 34.09,
|
124 |
+
"learning_rate": 9.187183650888056e-05,
|
125 |
+
"loss": 0.0174,
|
126 |
+
"step": 4500
|
127 |
+
},
|
128 |
+
{
|
129 |
+
"epoch": 34.09,
|
130 |
+
"eval_loss": 0.20308123528957367,
|
131 |
+
"eval_runtime": 964.2383,
|
132 |
+
"eval_samples_per_second": 5.364,
|
133 |
+
"eval_wer": 0.29469445388698584,
|
134 |
+
"step": 4500
|
135 |
}
|
136 |
],
|
137 |
+
"max_steps": 6600,
|
138 |
+
"num_train_epochs": 50,
|
139 |
+
"total_flos": 7.0460703447027696e+19,
|
140 |
"trial_name": null,
|
141 |
"trial_params": null
|
142 |
}
|
training_args.bin
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 2287
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:38479a919eda7dd28d3ae5f9cab083881c8741ef35462dbc9ab73b870d8c6d48
|
3 |
size 2287
|
vocab.json
CHANGED
@@ -1 +1 @@
|
|
1 |
-
{"
|
1 |
+
{"m": 0, "d": 1, "c": 2, "l": 3, "b": 4, "i": 5, "y": 6, "w": 7, "k": 8, "n": 10, "\u00f1": 11, "o": 12, "a": 13, "x": 14, "z": 15, "u": 16, "r": 17, "q": 18, "h": 19, "t": 20, "g": 21, "p": 22, "v": 23, "e": 24, "j": 25, "f": 26, "s": 27, "\u00ed": 28, "|": 9, "[UNK]": 29, "[PAD]": 30}
|