Hugging Face's logo

NX2411
/
wav2vec2-large-xlsr-korean-demo-test2

Automatic Speech Recognition
Transformers
PyTorch
TensorBoard
wav2vec2
Generated from Trainer
Inference Endpoints
Model card Files Metrics Community
1
NX2411 commited on
Commit
87dc428
·
1 Parent(s): fe97ea3

update model card README.md

Browse files
Files changed (1) hide show
  1. README.md +70 -53
README.md CHANGED
@@ -14,8 +14,8 @@ should probably proofread and complete it, then remove this comment. -->
14
 
15
  This model is a fine-tuned version of [facebook/wav2vec2-large-xlsr-53](https://huggingface.co/facebook/wav2vec2-large-xlsr-53) on the None dataset.
16
  It achieves the following results on the evaluation set:
17
- - Loss: 0.9948
18
- - Wer: 0.5865
19
 
20
  ## Model description
21
 
@@ -43,63 +43,80 @@ The following hyperparameters were used during training:
43
  - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
44
  - lr_scheduler_type: linear
45
  - lr_scheduler_warmup_steps: 500
46
- - num_epochs: 15
47
  - mixed_precision_training: Native AMP
48
 
49
  ### Training results
50
 
51
  | Training Loss | Epoch | Step | Validation Loss | Wer |
52
  |:-------------:|:-----:|:-----:|:---------------:|:------:|
53
- | 29.8545 | 0.3 | 400 | 5.3860 | 1.0 |
54
- | 4.9621 | 0.59 | 800 | 5.4067 | 1.0 |
55
- | 4.9254 | 0.89 | 1200 | 5.1930 | 1.0 |
56
- | 4.8425 | 1.19 | 1600 | 5.0176 | 1.0 |
57
- | 4.7955 | 1.49 | 2000 | 5.0994 | 1.0 |
58
- | 4.7091 | 1.78 | 2400 | 4.6204 | 1.0 |
59
- | 4.4177 | 2.08 | 2800 | 3.8672 | 1.0 |
60
- | 3.5708 | 2.38 | 3200 | 2.8938 | 0.9548 |
61
- | 2.9828 | 2.67 | 3600 | 2.4027 | 0.9100 |
62
- | 2.6781 | 2.97 | 4000 | 2.0710 | 0.8728 |
63
- | 2.3347 | 3.27 | 4400 | 1.8604 | 0.8474 |
64
- | 2.2081 | 3.57 | 4800 | 1.7831 | 0.8116 |
65
- | 2.1184 | 3.86 | 5200 | 1.6272 | 0.8012 |
66
- | 1.9834 | 4.16 | 5600 | 1.5311 | 0.8007 |
67
- | 1.8402 | 4.46 | 6000 | 1.4352 | 0.7659 |
68
- | 1.7859 | 4.75 | 6400 | 1.3503 | 0.7485 |
69
- | 1.7374 | 5.05 | 6800 | 1.3561 | 0.7674 |
70
- | 1.5966 | 5.35 | 7200 | 1.3319 | 0.7222 |
71
- | 1.5716 | 5.65 | 7600 | 1.2539 | 0.7112 |
72
- | 1.579 | 5.94 | 8000 | 1.2456 | 0.7028 |
73
- | 1.4429 | 6.24 | 8400 | 1.2081 | 0.6884 |
74
- | 1.4176 | 6.54 | 8800 | 1.1681 | 0.6914 |
75
- | 1.403 | 6.84 | 9200 | 1.1583 | 0.6874 |
76
- | 1.3417 | 7.13 | 9600 | 1.1235 | 0.6590 |
77
- | 1.267 | 7.43 | 10000 | 1.1538 | 0.6720 |
78
- | 1.268 | 7.73 | 10400 | 1.0878 | 0.6556 |
79
- | 1.2245 | 8.02 | 10800 | 1.0759 | 0.6347 |
80
- | 1.1437 | 8.32 | 11200 | 1.0815 | 0.6412 |
81
- | 1.1386 | 8.62 | 11600 | 1.1007 | 0.6352 |
82
- | 1.1045 | 8.92 | 12000 | 1.0574 | 0.6521 |
83
- | 1.0533 | 9.21 | 12400 | 1.0772 | 0.6332 |
84
- | 1.0274 | 9.51 | 12800 | 1.0622 | 0.6267 |
85
- | 1.0398 | 9.81 | 13200 | 1.0380 | 0.6322 |
86
- | 0.9869 | 10.1 | 13600 | 1.0654 | 0.6267 |
87
- | 0.9309 | 10.4 | 14000 | 1.0505 | 0.6153 |
88
- | 0.9231 | 10.7 | 14400 | 1.0300 | 0.6128 |
89
- | 0.9324 | 11.0 | 14800 | 0.9777 | 0.6098 |
90
- | 0.8467 | 11.29 | 15200 | 1.0123 | 0.6133 |
91
- | 0.8471 | 11.59 | 15600 | 1.0086 | 0.6014 |
92
- | 0.8601 | 11.89 | 16000 | 1.0051 | 0.6004 |
93
- | 0.8111 | 12.18 | 16400 | 1.0242 | 0.5994 |
94
- | 0.7525 | 12.48 | 16800 | 1.0015 | 0.5875 |
95
- | 0.7697 | 12.78 | 17200 | 0.9987 | 0.5954 |
96
- | 0.7585 | 13.08 | 17600 | 1.0040 | 0.5949 |
97
- | 0.7163 | 13.37 | 18000 | 0.9584 | 0.5895 |
98
- | 0.7041 | 13.67 | 18400 | 0.9795 | 0.5885 |
99
- | 0.7115 | 13.97 | 18800 | 0.9726 | 0.5840 |
100
- | 0.6907 | 14.26 | 19200 | 0.9809 | 0.5855 |
101
- | 0.6847 | 14.56 | 19600 | 0.9979 | 0.5870 |
102
- | 0.6641 | 14.86 | 20000 | 0.9948 | 0.5865 |
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
103
 
104
 
105
  ### Framework versions
 
14
 
15
  This model is a fine-tuned version of [facebook/wav2vec2-large-xlsr-53](https://huggingface.co/facebook/wav2vec2-large-xlsr-53) on the None dataset.
16
  It achieves the following results on the evaluation set:
17
+ - Loss: 1.0566
18
+ - Wer: 0.5224
19
 
20
  ## Model description
21
 
 
43
  - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
44
  - lr_scheduler_type: linear
45
  - lr_scheduler_warmup_steps: 500
46
+ - num_epochs: 20
47
  - mixed_precision_training: Native AMP
48
 
49
  ### Training results
50
 
51
  | Training Loss | Epoch | Step | Validation Loss | Wer |
52
  |:-------------:|:-----:|:-----:|:---------------:|:------:|
53
+ | 31.2541 | 0.3 | 400 | 5.4002 | 1.0 |
54
+ | 4.9419 | 0.59 | 800 | 5.3336 | 1.0 |
55
+ | 4.8926 | 0.89 | 1200 | 5.0531 | 1.0 |
56
+ | 4.7218 | 1.19 | 1600 | 4.5172 | 1.0 |
57
+ | 4.0218 | 1.49 | 2000 | 3.1418 | 0.9518 |
58
+ | 3.0654 | 1.78 | 2400 | 2.4376 | 0.9041 |
59
+ | 2.6226 | 2.08 | 2800 | 2.0151 | 0.8643 |
60
+ | 2.2944 | 2.38 | 3200 | 1.8025 | 0.8290 |
61
+ | 2.1872 | 2.67 | 3600 | 1.6469 | 0.7962 |
62
+ | 2.0747 | 2.97 | 4000 | 1.5165 | 0.7714 |
63
+ | 1.8479 | 3.27 | 4400 | 1.4281 | 0.7694 |
64
+ | 1.8288 | 3.57 | 4800 | 1.3791 | 0.7326 |
65
+ | 1.801 | 3.86 | 5200 | 1.3328 | 0.7177 |
66
+ | 1.6723 | 4.16 | 5600 | 1.2954 | 0.7192 |
67
+ | 1.5925 | 4.46 | 6000 | 1.3137 | 0.6953 |
68
+ | 1.5709 | 4.75 | 6400 | 1.2086 | 0.6973 |
69
+ | 1.5294 | 5.05 | 6800 | 1.1811 | 0.6730 |
70
+ | 1.3844 | 5.35 | 7200 | 1.2053 | 0.6769 |
71
+ | 1.3906 | 5.65 | 7600 | 1.1287 | 0.6556 |
72
+ | 1.4088 | 5.94 | 8000 | 1.1251 | 0.6466 |
73
+ | 1.2989 | 6.24 | 8400 | 1.1577 | 0.6546 |
74
+ | 1.2523 | 6.54 | 8800 | 1.0643 | 0.6377 |
75
+ | 1.2651 | 6.84 | 9200 | 1.0865 | 0.6417 |
76
+ | 1.2209 | 7.13 | 9600 | 1.0981 | 0.6272 |
77
+ | 1.1435 | 7.43 | 10000 | 1.1195 | 0.6317 |
78
+ | 1.1616 | 7.73 | 10400 | 1.0672 | 0.6327 |
79
+ | 1.1272 | 8.02 | 10800 | 1.0413 | 0.6248 |
80
+ | 1.043 | 8.32 | 11200 | 1.0555 | 0.6233 |
81
+ | 1.0523 | 8.62 | 11600 | 1.0372 | 0.6178 |
82
+ | 1.0208 | 8.92 | 12000 | 1.0170 | 0.6128 |
83
+ | 0.9895 | 9.21 | 12400 | 1.0354 | 0.5934 |
84
+ | 0.95 | 9.51 | 12800 | 1.1019 | 0.6039 |
85
+ | 0.9705 | 9.81 | 13200 | 1.0229 | 0.5855 |
86
+ | 0.9202 | 10.1 | 13600 | 1.0364 | 0.5919 |
87
+ | 0.8644 | 10.4 | 14000 | 1.0721 | 0.5984 |
88
+ | 0.8641 | 10.7 | 14400 | 1.0383 | 0.5905 |
89
+ | 0.8924 | 11.0 | 14800 | 0.9947 | 0.5760 |
90
+ | 0.7914 | 11.29 | 15200 | 1.0270 | 0.5885 |
91
+ | 0.7882 | 11.59 | 15600 | 1.0271 | 0.5741 |
92
+ | 0.8116 | 11.89 | 16000 | 0.9937 | 0.5741 |
93
+ | 0.7584 | 12.18 | 16400 | 0.9924 | 0.5626 |
94
+ | 0.7051 | 12.48 | 16800 | 1.0023 | 0.5572 |
95
+ | 0.7232 | 12.78 | 17200 | 1.0479 | 0.5512 |
96
+ | 0.7149 | 13.08 | 17600 | 1.0475 | 0.5765 |
97
+ | 0.6579 | 13.37 | 18000 | 1.0218 | 0.5552 |
98
+ | 0.6615 | 13.67 | 18400 | 1.0339 | 0.5631 |
99
+ | 0.6629 | 13.97 | 18800 | 1.0239 | 0.5621 |
100
+ | 0.6221 | 14.26 | 19200 | 1.0331 | 0.5537 |
101
+ | 0.6159 | 14.56 | 19600 | 1.0640 | 0.5532 |
102
+ | 0.6032 | 14.86 | 20000 | 1.0192 | 0.5567 |
103
+ | 0.5748 | 15.16 | 20400 | 1.0093 | 0.5507 |
104
+ | 0.5614 | 15.45 | 20800 | 1.0458 | 0.5472 |
105
+ | 0.5626 | 15.75 | 21200 | 1.0318 | 0.5398 |
106
+ | 0.5429 | 16.05 | 21600 | 1.0112 | 0.5278 |
107
+ | 0.5407 | 16.34 | 22000 | 1.0120 | 0.5278 |
108
+ | 0.511 | 16.64 | 22400 | 1.0335 | 0.5249 |
109
+ | 0.5316 | 16.94 | 22800 | 1.0146 | 0.5348 |
110
+ | 0.4949 | 17.24 | 23200 | 1.0287 | 0.5388 |
111
+ | 0.496 | 17.53 | 23600 | 1.0229 | 0.5348 |
112
+ | 0.4986 | 17.83 | 24000 | 1.0094 | 0.5313 |
113
+ | 0.4787 | 18.13 | 24400 | 1.0620 | 0.5234 |
114
+ | 0.4508 | 18.42 | 24800 | 1.0401 | 0.5323 |
115
+ | 0.4754 | 18.72 | 25200 | 1.0543 | 0.5303 |
116
+ | 0.4584 | 19.02 | 25600 | 1.0433 | 0.5194 |
117
+ | 0.4431 | 19.32 | 26000 | 1.0597 | 0.5249 |
118
+ | 0.4448 | 19.61 | 26400 | 1.0548 | 0.5229 |
119
+ | 0.4475 | 19.91 | 26800 | 1.0566 | 0.5224 |
120
 
121
 
122
  ### Framework versions