Hugging Face's logo

AndrewMcDowell
/
wav2vec2-xls-r-1B-german

Automatic Speech Recognition
Transformers
PyTorch
German
wav2vec2
mozilla-foundation/common_voice_8_0
Generated from Trainer
robust-speech-event
hf-asr-leaderboard
Eval Results
Inference Endpoints
Model card Files Community
AndrewMcDowell commited on
Commit
97f1966
•
1 Parent(s): a061199

update model card README.md

Browse files
Files changed (1) hide show
  1. README.md +40 -74
README.md CHANGED
@@ -1,46 +1,12 @@
1
  ---
2
- language:
3
- - de
4
  license: apache-2.0
5
  tags:
6
- - automatic-speech-recognition
7
- - mozilla-foundation/common_voice_8_0
8
  - generated_from_trainer
9
- - robust-speech-event
10
- - de
11
  datasets:
12
  - common_voice
13
  model-index:
14
- - name: 'XLS-R-1B - German'
15
- results:
16
- - task:
17
- name: Automatic Speech Recognition
18
- type: automatic-speech-recognition
19
- dataset:
20
- name: Common Voice 8
21
- type: mozilla-foundation/common_voice_8_0
22
- args: de
23
- metrics:
24
- - name: Test WER
25
- type: wer
26
- value: 16.12
27
- - name: Test CER
28
- type: cer
29
- value: 4.023836120707133
30
- - task:
31
- name: Automatic Speech Recognition
32
- type: automatic-speech-recognition
33
- dataset:
34
- name: Robust Speech Event - Dev Data
35
- type: speech-recognition-community-v2/dev_data
36
- args: de
37
- metrics:
38
- - name: Test WER
39
- type: wer
40
- value: 36.76
41
- - name: Test CER
42
- type: cer
43
- value: 14.03
44
  ---
45
 
46
  <!-- This model card has been generated automatically according to the information the Trainer had access to. You
@@ -48,10 +14,10 @@ should probably proofread and complete it, then remove this comment. -->
48
 
49
  #
50
 
51
- This model is a fine-tuned version of [facebook/wav2vec2-xls-r-1b](https://huggingface.co/facebook/wav2vec2-xls-r-1b) on the MOZILLA-FOUNDATION/COMMON_VOICE_8_0 - DE dataset.
52
  It achieves the following results on the evaluation set:
53
- - Loss: 0.1476
54
- - Wer: 0.1612
55
 
56
  ## Model description
57
 
@@ -79,40 +45,47 @@ The following hyperparameters were used during training:
79
  - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
80
  - lr_scheduler_type: linear
81
  - lr_scheduler_warmup_steps: 2000
82
- - num_epochs: 2.0
83
  - mixed_precision_training: Native AMP
84
 
85
  ### Training results
86
 
87
  | Training Loss | Epoch | Step | Validation Loss | Wer |
88
  |:-------------:|:-----:|:-----:|:---------------:|:------:|
89
- | 1.1842 | 0.07 | 1000 | 0.4461 | 0.4918 |
90
- | 1.1317 | 0.15 | 2000 | 0.2669 | 0.2748 |
91
- | 1.1029 | 0.22 | 3000 | 0.2638 | 0.2706 |
92
- | 1.0949 | 0.29 | 4000 | 0.2519 | 0.2627 |
93
- | 1.0923 | 0.37 | 5000 | 0.2475 | 0.2577 |
94
- | 1.0847 | 0.44 | 6000 | 0.2436 | 0.2612 |
95
- | 1.0667 | 0.51 | 7000 | 0.2472 | 0.2661 |
96
- | 1.0709 | 0.59 | 8000 | 0.2489 | 0.2610 |
97
- | 1.0472 | 0.66 | 9000 | 0.2354 | 0.2500 |
98
- | 1.0604 | 0.73 | 10000 | 0.2346 | 0.2485 |
99
- | 1.0375 | 0.81 | 11000 | 0.2286 | 0.2390 |
100
- | 1.0193 | 0.88 | 12000 | 0.2212 | 0.2338 |
101
- | 1.0077 | 0.95 | 13000 | 0.2152 | 0.2269 |
102
- | 1.0004 | 1.03 | 14000 | 0.2093 | 0.2207 |
103
- | 0.9649 | 1.1 | 15000 | 0.1993 | 0.2113 |
104
- | 0.9509 | 1.17 | 16000 | 0.1934 | 0.2089 |
105
- | 0.9533 | 1.25 | 17000 | 0.1874 | 0.2023 |
106
- | 0.9248 | 1.32 | 18000 | 0.1818 | 0.1974 |
107
- | 0.9216 | 1.39 | 19000 | 0.1776 | 0.1926 |
108
- | 0.8964 | 1.47 | 20000 | 0.1722 | 0.1904 |
109
- | 0.8941 | 1.54 | 21000 | 0.1690 | 0.1852 |
110
- | 0.871 | 1.61 | 22000 | 0.1627 | 0.1781 |
111
- | 0.847 | 1.69 | 23000 | 0.1591 | 0.1751 |
112
- | 0.822 | 1.76 | 24000 | 0.1551 | 0.1701 |
113
- | 0.8188 | 1.83 | 25000 | 0.1528 | 0.1667 |
114
- | 0.8305 | 1.91 | 26000 | 0.1492 | 0.1631 |
115
- | 0.8122 | 1.98 | 27000 | 0.1479 | 0.1611 |
 
 
 
 
 
 
 
116
 
117
 
118
  ### Framework versions
@@ -121,10 +94,3 @@ The following hyperparameters were used during training:
121
  - Pytorch 1.10.2+cu102
122
  - Datasets 1.18.2.dev0
123
  - Tokenizers 0.11.0
124
-
125
- #### Evaluation Commands
126
- 1. To evaluate on `mozilla-foundation/common_voice_8_0` with split `test`
127
-
128
- ```bash
129
- python ./eval.py --model_id AndrewMcDowell/wav2vec2-xls-r-1B-german-de --dataset mozilla-foundation/common_voice_8_0 --config de --split test --log_outputs
130
- ```
 
1
  ---
 
 
2
  license: apache-2.0
3
  tags:
 
 
4
  - generated_from_trainer
 
 
5
  datasets:
6
  - common_voice
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
 
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 common_voice dataset.
18
  It achieves the following results on the evaluation set:
19
+ - Loss: 0.1355
20
+ - Wer: 0.1531
21
 
22
  ## Model description
23
 
 
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: 2.5
49
  - mixed_precision_training: Native AMP
50
 
51
  ### Training results
52
 
53
  | Training Loss | Epoch | Step | Validation Loss | Wer |
54
  |:-------------:|:-----:|:-----:|:---------------:|:------:|
55
+ | 1.0826 | 0.07 | 1000 | 0.4637 | 0.4654 |
56
+ | 1.118 | 0.15 | 2000 | 0.2595 | 0.2687 |
57
+ | 1.1268 | 0.22 | 3000 | 0.2635 | 0.2661 |
58
+ | 1.0919 | 0.29 | 4000 | 0.2417 | 0.2566 |
59
+ | 1.1013 | 0.37 | 5000 | 0.2414 | 0.2567 |
60
+ | 1.0898 | 0.44 | 6000 | 0.2546 | 0.2731 |
61
+ | 1.0808 | 0.51 | 7000 | 0.2399 | 0.2535 |
62
+ | 1.0719 | 0.59 | 8000 | 0.2353 | 0.2528 |
63
+ | 1.0446 | 0.66 | 9000 | 0.2427 | 0.2545 |
64
+ | 1.0347 | 0.73 | 10000 | 0.2266 | 0.2402 |
65
+ | 1.0457 | 0.81 | 11000 | 0.2290 | 0.2448 |
66
+ | 1.0124 | 0.88 | 12000 | 0.2295 | 0.2448 |
67
+ | 1.025 | 0.95 | 13000 | 0.2138 | 0.2345 |
68
+ | 1.0107 | 1.03 | 14000 | 0.2108 | 0.2294 |
69
+ | 0.9758 | 1.1 | 15000 | 0.2019 | 0.2204 |
70
+ | 0.9547 | 1.17 | 16000 | 0.2000 | 0.2178 |
71
+ | 0.986 | 1.25 | 17000 | 0.2018 | 0.2200 |
72
+ | 0.9588 | 1.32 | 18000 | 0.1992 | 0.2138 |
73
+ | 0.9413 | 1.39 | 19000 | 0.1898 | 0.2049 |
74
+ | 0.9339 | 1.47 | 20000 | 0.1874 | 0.2056 |
75
+ | 0.9268 | 1.54 | 21000 | 0.1797 | 0.1976 |
76
+ | 0.9194 | 1.61 | 22000 | 0.1743 | 0.1905 |
77
+ | 0.8987 | 1.69 | 23000 | 0.1738 | 0.1932 |
78
+ | 0.8884 | 1.76 | 24000 | 0.1703 | 0.1873 |
79
+ | 0.8939 | 1.83 | 25000 | 0.1633 | 0.1831 |
80
+ | 0.8629 | 1.91 | 26000 | 0.1549 | 0.1750 |
81
+ | 0.8607 | 1.98 | 27000 | 0.1550 | 0.1738 |
82
+ | 0.8316 | 2.05 | 28000 | 0.1512 | 0.1709 |
83
+ | 0.8321 | 2.13 | 29000 | 0.1481 | 0.1657 |
84
+ | 0.825 | 2.2 | 30000 | 0.1446 | 0.1627 |
85
+ | 0.8115 | 2.27 | 31000 | 0.1396 | 0.1583 |
86
+ | 0.7959 | 2.35 | 32000 | 0.1389 | 0.1569 |
87
+ | 0.7835 | 2.42 | 33000 | 0.1362 | 0.1545 |
88
+ | 0.7959 | 2.49 | 34000 | 0.1355 | 0.1531 |
89
 
90
 
91
  ### Framework versions
 
94
  - Pytorch 1.10.2+cu102
95
  - Datasets 1.18.2.dev0
96
  - Tokenizers 0.11.0