kingabzpro commited on
Commit
d092371
1 Parent(s): 270ecdd

update model card README.md

Browse files
Files changed (1) hide show
  1. README.md +20 -64
README.md CHANGED
@@ -1,57 +1,12 @@
1
  ---
2
- language:
3
- - sv-SE
4
-
5
  license: apache-2.0
6
  tags:
7
- - automatic-speech-recognition
8
- - robust-speech-event
9
  datasets:
10
- - mozilla-foundation/common_voice_8_0
11
- metrics:
12
- - wer
13
- - cer
14
  model-index:
15
  - name: wav2vec2-large-xls-r-1b-Swedish
16
- results:
17
- - task:
18
- type: automatic-speech-recognition # Required. Example: automatic-speech-recognition
19
- name: Speech Recognition # Optional. Example: Speech Recognition
20
- dataset:
21
- type: mozilla-foundation/common_voice_8_0 # Required. Example: common_voice. Use dataset id from https://hf.co/datasets
22
- name: Common Voice sv-SE # Required. Example: Common Voice zh-CN
23
- args: sv-SE # Optional. Example: zh-CN
24
- metrics:
25
- - type: wer # Required. Example: wer
26
- value: 18.03 # Required. Example: 20.90
27
- name: Test WER Without LM # Optional. Example: Test WER
28
- args:
29
- - learning_rate: 7.5e-05
30
- - train_batch_size: 32
31
- - eval_batch_size: 8
32
- - seed: 42
33
- - gradient_accumulation_steps: 4
34
- - total_train_batch_size: 128
35
- - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
36
- - lr_scheduler_type: linear
37
- - lr_scheduler_warmup_steps: 1000
38
- - num_epochs: 50
39
- - mixed_precision_training: Native AMP # Optional. Example for BLEU: max_order
40
- - type: cer # Required. Example: wer
41
- value: 5.69 # Required. Example: 20.90
42
- name: Test CER Without LM # Optional. Example: Test WER
43
- args:
44
- - learning_rate: 7.5e-05
45
- - train_batch_size: 32
46
- - eval_batch_size: 8
47
- - seed: 42
48
- - gradient_accumulation_steps: 4
49
- - total_train_batch_size: 128
50
- - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
51
- - lr_scheduler_type: linear
52
- - lr_scheduler_warmup_steps: 1000
53
- - num_epochs: 50
54
- - mixed_precision_training: Native AMP
55
  ---
56
 
57
  <!-- This model card has been generated automatically according to the information the Trainer had access to. You
@@ -61,15 +16,21 @@ should probably proofread and complete it, then remove this comment. -->
61
 
62
  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.
63
  It achieves the following results on the evaluation set:
 
 
 
 
 
 
 
64
 
65
- **Without LM**
66
- - Loss: 0.3370
67
- - Wer: 0.1803
68
- - Cer: 0.0569
69
 
70
- **With LM**
71
 
 
72
 
 
73
 
74
  ## Training procedure
75
 
@@ -77,11 +38,11 @@ It achieves the following results on the evaluation set:
77
 
78
  The following hyperparameters were used during training:
79
  - learning_rate: 7.5e-05
80
- - train_batch_size: 32
81
  - eval_batch_size: 8
82
  - seed: 42
83
  - gradient_accumulation_steps: 4
84
- - total_train_batch_size: 128
85
  - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
86
  - lr_scheduler_type: linear
87
  - lr_scheduler_warmup_steps: 1000
@@ -92,15 +53,10 @@ The following hyperparameters were used during training:
92
 
93
  | Training Loss | Epoch | Step | Validation Loss | Wer | Cer |
94
  |:-------------:|:-----:|:----:|:---------------:|:------:|:------:|
95
- | 3.1423 | 5.49 | 500 | 0.5523 | 0.4414 | 0.1313 |
96
- | 0.8615 | 10.98 | 1000 | 0.3877 | 0.2946 | 0.0942 |
97
- | 0.4848 | 16.48 | 1500 | 0.3580 | 0.2539 | 0.0798 |
98
- | 0.3538 | 21.97 | 2000 | 0.3391 | 0.2254 | 0.0709 |
99
- | 0.2879 | 27.47 | 2500 | 0.3392 | 0.2151 | 0.0680 |
100
- | 0.2466 | 32.96 | 3000 | 0.3687 | 0.2131 | 0.0680 |
101
- | 0.2146 | 38.46 | 3500 | 0.3551 | 0.1951 | 0.0618 |
102
- | 0.1916 | 43.95 | 4000 | 0.3601 | 0.1867 | 0.0590 |
103
- | 0.175 | 49.45 | 4500 | 0.3370 | 0.1803 | 0.0569 |
104
 
105
 
106
  ### Framework versions
 
1
  ---
 
 
 
2
  license: apache-2.0
3
  tags:
4
+ - generated_from_trainer
 
5
  datasets:
6
+ - common_voice
 
 
 
7
  model-index:
8
  - name: wav2vec2-large-xls-r-1b-Swedish
9
+ results: []
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
10
  ---
11
 
12
  <!-- This model card has been generated automatically according to the information the Trainer had access to. You
 
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: 0.3232
20
+ - Wer: 0.1844
21
+ - Cer: 0.0575
22
+
23
+ ## Model description
24
+
25
+ More information needed
26
 
27
+ ## Intended uses & limitations
 
 
 
28
 
29
+ More information needed
30
 
31
+ ## Training and evaluation data
32
 
33
+ More information needed
34
 
35
  ## Training procedure
36
 
 
38
 
39
  The following hyperparameters were used during training:
40
  - learning_rate: 7.5e-05
41
+ - train_batch_size: 64
42
  - eval_batch_size: 8
43
  - seed: 42
44
  - gradient_accumulation_steps: 4
45
+ - total_train_batch_size: 256
46
  - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
47
  - lr_scheduler_type: linear
48
  - lr_scheduler_warmup_steps: 1000
 
53
 
54
  | Training Loss | Epoch | Step | Validation Loss | Wer | Cer |
55
  |:-------------:|:-----:|:----:|:---------------:|:------:|:------:|
56
+ | 3.1562 | 11.11 | 500 | 0.4830 | 0.3729 | 0.1169 |
57
+ | 0.5655 | 22.22 | 1000 | 0.3553 | 0.2381 | 0.0743 |
58
+ | 0.3376 | 33.33 | 1500 | 0.3359 | 0.2179 | 0.0696 |
59
+ | 0.2419 | 44.44 | 2000 | 0.3232 | 0.1844 | 0.0575 |
 
 
 
 
 
60
 
61
 
62
  ### Framework versions