Hugging Face's logo

oliverguhr
/
wav2vec2-base-german-cv9

Automatic Speech Recognition
Transformers
PyTorch
TensorBoard
Safetensors
German
wav2vec2
mozilla-foundation/common_voice_9_0
Generated from Trainer
Eval Results
Inference Endpoints
Model card Files Metrics Community
2
oliverguhr commited on
Commit
62829c3
·
1 Parent(s): ea81c28

set license

Browse files
Files changed (1) hide show
  1. README.md +142 -0
README.md ADDED
@@ -0,0 +1,142 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ language:
3
+ - de
4
+ license: mit
5
+ tags:
6
+ - automatic-speech-recognition
7
+ - mozilla-foundation/common_voice_9_0
8
+ - generated_from_trainer
9
+ datasets:
10
+ - mozilla-foundation/common_voice_9_0
11
+ model-index:
12
+ - name: wav2vec2-base-german-cv9
13
+ results:
14
+ - task:
15
+ name: Automatic Speech Recognition
16
+ type: automatic-speech-recognition
17
+ dataset:
18
+ name: Common Voice 6.1
19
+ type: common_voice
20
+ args: de
21
+ metrics:
22
+ - name: Test WER
23
+ type: wer
24
+ value: 10.565782902002716
25
+ - name: Test CER
26
+ type: cer
27
+ value: 2.6226824852959657
28
+ - task:
29
+ name: Automatic Speech Recognition
30
+ type: automatic-speech-recognition
31
+ dataset:
32
+ name: Common Voice 6.1
33
+ type: common_voice
34
+ args: de
35
+ metrics:
36
+ - name: Test WER (+LM)
37
+ type: wer
38
+ value: 7.996088831362508
39
+ - name: Test CER (+LM)
40
+ type: cer
41
+ value: 2.1515717711623326
42
+ ---
43
+
44
+
45
+ # wav2vec2-base-german-cv9
46
+
47
+ This model is a fine-tuned version of [facebook/wav2vec2-base](https://huggingface.co/facebook/wav2vec2-base) on the MOZILLA-FOUNDATION/COMMON_VOICE_9_0 - DE dataset.
48
+ It achieves the following results on the evaluation set:
49
+ - Loss: 0.1742
50
+ - Wer: 0.1209
51
+
52
+ ## Model description
53
+
54
+ More information needed
55
+
56
+ ## Intended uses & limitations
57
+
58
+ More information needed
59
+
60
+ ## Training and evaluation data
61
+
62
+ More information needed
63
+
64
+ ## Training procedure
65
+
66
+ ### Training hyperparameters
67
+
68
+ The following hyperparameters were used during training:
69
+ - learning_rate: 0.0001
70
+ - train_batch_size: 16
71
+ - eval_batch_size: 32
72
+ - seed: 42
73
+ - gradient_accumulation_steps: 8
74
+ - total_train_batch_size: 128
75
+ - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
76
+ - lr_scheduler_type: linear
77
+ - lr_scheduler_warmup_ratio: 0.1
78
+ - num_epochs: 50.0
79
+ - mixed_precision_training: Native AMP
80
+
81
+ ### Training results
82
+
83
+ | Training Loss | Epoch | Step | Validation Loss | Wer |
84
+ |:-------------:|:-----:|:------:|:---------------:|:------:|
85
+ | 0.6827 | 1.0 | 3557 | 0.6695 | 0.6247 |
86
+ | 0.3992 | 2.0 | 7114 | 0.3738 | 0.3936 |
87
+ | 0.2611 | 3.0 | 10671 | 0.3011 | 0.3177 |
88
+ | 0.2536 | 4.0 | 14228 | 0.2672 | 0.2749 |
89
+ | 0.1943 | 5.0 | 17785 | 0.2487 | 0.2480 |
90
+ | 0.2004 | 6.0 | 21342 | 0.2246 | 0.2268 |
91
+ | 0.1605 | 7.0 | 24899 | 0.2176 | 0.2120 |
92
+ | 0.1579 | 8.0 | 28456 | 0.2046 | 0.2024 |
93
+ | 0.1668 | 9.0 | 32013 | 0.2027 | 0.1944 |
94
+ | 0.1338 | 10.0 | 35570 | 0.1968 | 0.1854 |
95
+ | 0.1478 | 11.0 | 39127 | 0.1963 | 0.1823 |
96
+ | 0.1177 | 12.0 | 42684 | 0.1956 | 0.1800 |
97
+ | 0.1245 | 13.0 | 46241 | 0.1889 | 0.1732 |
98
+ | 0.1124 | 14.0 | 49798 | 0.1868 | 0.1714 |
99
+ | 0.1112 | 15.0 | 53355 | 0.1805 | 0.1650 |
100
+ | 0.1209 | 16.0 | 56912 | 0.1860 | 0.1614 |
101
+ | 0.1002 | 17.0 | 60469 | 0.1828 | 0.1604 |
102
+ | 0.118 | 18.0 | 64026 | 0.1832 | 0.1580 |
103
+ | 0.0974 | 19.0 | 67583 | 0.1771 | 0.1555 |
104
+ | 0.1007 | 20.0 | 71140 | 0.1812 | 0.1532 |
105
+ | 0.0866 | 21.0 | 74697 | 0.1752 | 0.1504 |
106
+ | 0.0901 | 22.0 | 78254 | 0.1690 | 0.1477 |
107
+ | 0.0964 | 23.0 | 81811 | 0.1773 | 0.1489 |
108
+ | 0.085 | 24.0 | 85368 | 0.1776 | 0.1456 |
109
+ | 0.0945 | 25.0 | 88925 | 0.1786 | 0.1428 |
110
+ | 0.0804 | 26.0 | 92482 | 0.1737 | 0.1429 |
111
+ | 0.0832 | 27.0 | 96039 | 0.1789 | 0.1394 |
112
+ | 0.0683 | 28.0 | 99596 | 0.1741 | 0.1390 |
113
+ | 0.0761 | 29.0 | 103153 | 0.1688 | 0.1379 |
114
+ | 0.0833 | 30.0 | 106710 | 0.1726 | 0.1370 |
115
+ | 0.0753 | 31.0 | 110267 | 0.1774 | 0.1353 |
116
+ | 0.08 | 32.0 | 113824 | 0.1734 | 0.1344 |
117
+ | 0.0644 | 33.0 | 117381 | 0.1737 | 0.1334 |
118
+ | 0.0745 | 34.0 | 120938 | 0.1763 | 0.1335 |
119
+ | 0.0629 | 35.0 | 124495 | 0.1761 | 0.1311 |
120
+ | 0.0654 | 36.0 | 128052 | 0.1718 | 0.1302 |
121
+ | 0.0656 | 37.0 | 131609 | 0.1697 | 0.1301 |
122
+ | 0.0643 | 38.0 | 135166 | 0.1716 | 0.1279 |
123
+ | 0.0683 | 39.0 | 138723 | 0.1777 | 0.1279 |
124
+ | 0.0587 | 40.0 | 142280 | 0.1735 | 0.1271 |
125
+ | 0.0693 | 41.0 | 145837 | 0.1780 | 0.1260 |
126
+ | 0.0532 | 42.0 | 149394 | 0.1724 | 0.1245 |
127
+ | 0.0594 | 43.0 | 152951 | 0.1736 | 0.1250 |
128
+ | 0.0544 | 44.0 | 156508 | 0.1744 | 0.1238 |
129
+ | 0.0559 | 45.0 | 160065 | 0.1770 | 0.1232 |
130
+ | 0.0557 | 46.0 | 163622 | 0.1766 | 0.1231 |
131
+ | 0.0521 | 47.0 | 167179 | 0.1751 | 0.1220 |
132
+ | 0.0591 | 48.0 | 170736 | 0.1724 | 0.1217 |
133
+ | 0.0507 | 49.0 | 174293 | 0.1753 | 0.1212 |
134
+ | 0.0577 | 50.0 | 177850 | 0.1742 | 0.1209 |
135
+
136
+
137
+ ### Framework versions
138
+
139
+ - Transformers 4.20.1
140
+ - Pytorch 1.11.0+cu113
141
+ - Datasets 2.0.0
142
+ - Tokenizers 0.11.6