Hugging Face's logo

anantoj
/
wav2vec2-xls-r-1b-korean

Automatic Speech Recognition
Transformers
PyTorch
Korean
wav2vec2
Generated from Trainer
hf-asr-leaderboard
robust-speech-event
Eval Results
Inference Endpoints
Model card Files Community
2
anantoj commited on
Commit
de79e0d
•
1 Parent(s): 3233080

update model card README.md

Browse files
Files changed (2) hide show
  1. README.md +131 -0
  2. nohup.out +17 -0
README.md ADDED
@@ -0,0 +1,131 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ tags:
4
+ - generated_from_trainer
5
+ datasets:
6
+ - zeroth_korean_asr
7
+ model-index:
8
+ - name: ''
9
+ results: []
10
+ ---
11
+
12
+ <!-- This model card has been generated automatically according to the information the Trainer had access to. You
13
+ should probably proofread and complete it, then remove this comment. -->
14
+
15
+ #
16
+
17
+ This model is a fine-tuned version of [facebook/wav2vec2-xls-r-1b](https://huggingface.co/facebook/wav2vec2-xls-r-1b) on the zeroth_korean_asr dataset.
18
+ It achieves the following results on the evaluation set:
19
+ - Loss: 0.0643
20
+ - Wer: 0.0455
21
+
22
+ ## Model description
23
+
24
+ More information needed
25
+
26
+ ## Intended uses & limitations
27
+
28
+ More information needed
29
+
30
+ ## Training and evaluation data
31
+
32
+ More information needed
33
+
34
+ ## Training procedure
35
+
36
+ ### Training hyperparameters
37
+
38
+ The following hyperparameters were used during training:
39
+ - learning_rate: 7.5e-05
40
+ - train_batch_size: 8
41
+ - eval_batch_size: 8
42
+ - seed: 42
43
+ - gradient_accumulation_steps: 4
44
+ - total_train_batch_size: 32
45
+ - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
46
+ - lr_scheduler_type: linear
47
+ - lr_scheduler_warmup_steps: 2000
48
+ - num_epochs: 50.0
49
+ - mixed_precision_training: Native AMP
50
+
51
+ ### Training results
52
+
53
+ | Training Loss | Epoch | Step | Validation Loss | Wer |
54
+ |:-------------:|:-----:|:-----:|:---------------:|:------:|
55
+ | 4.603 | 0.72 | 500 | 4.6572 | 0.9985 |
56
+ | 2.6314 | 1.44 | 1000 | 2.0424 | 0.9256 |
57
+ | 2.2708 | 2.16 | 1500 | 0.9889 | 0.6989 |
58
+ | 2.1769 | 2.88 | 2000 | 0.8366 | 0.6312 |
59
+ | 2.1142 | 3.6 | 2500 | 0.7555 | 0.5998 |
60
+ | 2.0084 | 4.32 | 3000 | 0.7144 | 0.6003 |
61
+ | 1.9272 | 5.04 | 3500 | 0.6311 | 0.5461 |
62
+ | 1.8687 | 5.75 | 4000 | 0.6252 | 0.5430 |
63
+ | 1.8186 | 6.47 | 4500 | 0.5491 | 0.4988 |
64
+ | 1.7364 | 7.19 | 5000 | 0.5463 | 0.4959 |
65
+ | 1.6809 | 7.91 | 5500 | 0.4724 | 0.4484 |
66
+ | 1.641 | 8.63 | 6000 | 0.4679 | 0.4461 |
67
+ | 1.572 | 9.35 | 6500 | 0.4387 | 0.4236 |
68
+ | 1.5256 | 10.07 | 7000 | 0.3970 | 0.4003 |
69
+ | 1.5044 | 10.79 | 7500 | 0.3690 | 0.3893 |
70
+ | 1.4563 | 11.51 | 8000 | 0.3752 | 0.3875 |
71
+ | 1.394 | 12.23 | 8500 | 0.3386 | 0.3567 |
72
+ | 1.3641 | 12.95 | 9000 | 0.3290 | 0.3467 |
73
+ | 1.2878 | 13.67 | 9500 | 0.2893 | 0.3135 |
74
+ | 1.2602 | 14.39 | 10000 | 0.2723 | 0.3029 |
75
+ | 1.2302 | 15.11 | 10500 | 0.2603 | 0.2989 |
76
+ | 1.1865 | 15.83 | 11000 | 0.2440 | 0.2794 |
77
+ | 1.1491 | 16.55 | 11500 | 0.2500 | 0.2788 |
78
+ | 1.093 | 17.27 | 12000 | 0.2279 | 0.2629 |
79
+ | 1.0367 | 17.98 | 12500 | 0.2076 | 0.2443 |
80
+ | 0.9954 | 18.7 | 13000 | 0.1844 | 0.2259 |
81
+ | 0.99 | 19.42 | 13500 | 0.1794 | 0.2179 |
82
+ | 0.9385 | 20.14 | 14000 | 0.1765 | 0.2122 |
83
+ | 0.8952 | 20.86 | 14500 | 0.1706 | 0.1974 |
84
+ | 0.8841 | 21.58 | 15000 | 0.1791 | 0.1969 |
85
+ | 0.847 | 22.3 | 15500 | 0.1780 | 0.2060 |
86
+ | 0.8669 | 23.02 | 16000 | 0.1608 | 0.1862 |
87
+ | 0.8066 | 23.74 | 16500 | 0.1447 | 0.1626 |
88
+ | 0.7908 | 24.46 | 17000 | 0.1457 | 0.1655 |
89
+ | 0.7459 | 25.18 | 17500 | 0.1350 | 0.1445 |
90
+ | 0.7218 | 25.9 | 18000 | 0.1276 | 0.1421 |
91
+ | 0.703 | 26.62 | 18500 | 0.1177 | 0.1302 |
92
+ | 0.685 | 27.34 | 19000 | 0.1147 | 0.1305 |
93
+ | 0.6811 | 28.06 | 19500 | 0.1128 | 0.1244 |
94
+ | 0.6444 | 28.78 | 20000 | 0.1120 | 0.1213 |
95
+ | 0.6323 | 29.5 | 20500 | 0.1137 | 0.1166 |
96
+ | 0.5998 | 30.22 | 21000 | 0.1051 | 0.1107 |
97
+ | 0.5706 | 30.93 | 21500 | 0.1035 | 0.1037 |
98
+ | 0.5555 | 31.65 | 22000 | 0.1031 | 0.0927 |
99
+ | 0.5389 | 32.37 | 22500 | 0.0997 | 0.0900 |
100
+ | 0.5201 | 33.09 | 23000 | 0.0920 | 0.0912 |
101
+ | 0.5146 | 33.81 | 23500 | 0.0929 | 0.0947 |
102
+ | 0.515 | 34.53 | 24000 | 0.1000 | 0.0953 |
103
+ | 0.4743 | 35.25 | 24500 | 0.0922 | 0.0892 |
104
+ | 0.4707 | 35.97 | 25000 | 0.0852 | 0.0808 |
105
+ | 0.4456 | 36.69 | 25500 | 0.0855 | 0.0779 |
106
+ | 0.443 | 37.41 | 26000 | 0.0843 | 0.0738 |
107
+ | 0.4388 | 38.13 | 26500 | 0.0816 | 0.0699 |
108
+ | 0.4162 | 38.85 | 27000 | 0.0752 | 0.0645 |
109
+ | 0.3979 | 39.57 | 27500 | 0.0761 | 0.0621 |
110
+ | 0.3889 | 40.29 | 28000 | 0.0771 | 0.0625 |
111
+ | 0.3923 | 41.01 | 28500 | 0.0755 | 0.0598 |
112
+ | 0.3693 | 41.73 | 29000 | 0.0730 | 0.0578 |
113
+ | 0.3642 | 42.45 | 29500 | 0.0739 | 0.0598 |
114
+ | 0.3532 | 43.17 | 30000 | 0.0712 | 0.0553 |
115
+ | 0.3513 | 43.88 | 30500 | 0.0762 | 0.0516 |
116
+ | 0.3349 | 44.6 | 31000 | 0.0731 | 0.0504 |
117
+ | 0.3305 | 45.32 | 31500 | 0.0725 | 0.0507 |
118
+ | 0.3285 | 46.04 | 32000 | 0.0709 | 0.0489 |
119
+ | 0.3179 | 46.76 | 32500 | 0.0667 | 0.0467 |
120
+ | 0.3158 | 47.48 | 33000 | 0.0653 | 0.0494 |
121
+ | 0.3033 | 48.2 | 33500 | 0.0638 | 0.0456 |
122
+ | 0.3023 | 48.92 | 34000 | 0.0644 | 0.0464 |
123
+ | 0.2975 | 49.64 | 34500 | 0.0643 | 0.0455 |
124
+
125
+
126
+ ### Framework versions
127
+
128
+ - Transformers 4.17.0.dev0
129
+ - Pytorch 1.10.2+cu102
130
+ - Datasets 1.18.3.dev0
131
+ - Tokenizers 0.11.0
nohup.out CHANGED
@@ -5696,3 +5696,20 @@ Saving model checkpoint to ./
5696
  Configuration saved in ./config.json
5697
  Model weights saved in ./pytorch_model.bin
5698
  Configuration saved in ./preprocessor_config.json
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
5696
  Configuration saved in ./config.json
5697
  Model weights saved in ./pytorch_model.bin
5698
  Configuration saved in ./preprocessor_config.json
5699
+ Several commits (2) will be pushed upstream.
5700
+ {'eval_loss': 0.06428828090429306, 'eval_wer': 0.045454545454545456, 'eval_runtime': 56.8585, 'eval_samples_per_second': 8.02, 'eval_steps_per_second': 1.002, 'epoch': 49.64}
5701
+ {'loss': 0.2955, 'learning_rate': 3.8702290076335876e-07, 'epoch': 49.78}
5702
+ {'loss': 0.2952, 'learning_rate': 1.580152671755725e-07, 'epoch': 49.93}
5703
+ {'train_runtime': 194988.743, 'train_samples_per_second': 5.709, 'train_steps_per_second': 0.178, 'train_loss': 1.0736439298451375, 'epoch': 50.0}
5704
+ 02/02/2022 17:36:28 - WARNING - huggingface_hub.repository - Several commits (2) will be pushed upstream.
5705
+ The progress bars may be unreliable.
5706
+ 02/02/2022 17:36:28 - WARNING - huggingface_hub.repository - The progress bars may be unreliable.
5707
+
5708
+ ac4fd85..3233080 main -> main
5709
+
5710
+ 02/02/2022 17:38:39 - WARNING - huggingface_hub.repository - To https://huggingface.co/anantoj/wav2vec2-xls-r-1b-korean
5711
+ ac4fd85..3233080 main -> main
5712
+
5713
+
5714
+ Dropping the following result as it does not have all the necessary fields:
5715
+ {'dataset': {'name': 'zeroth_korean_asr', 'type': 'zeroth_korean_asr', 'args': 'clean'}}