Hugging Face's logo

gary109
/
ai-light-dance_singing_ft_pretrain_wav2vec2-large-lv60

Automatic Speech Recognition
Transformers
PyTorch
TensorBoard
wav2vec2
gary109/AI_Light_Dance
generated_from_trainer
Inference Endpoints
Model card Files Metrics Community
1
gary109 commited on
Commit
8eabdee
•
1 Parent(s): e9e5ab3

update model card README.md

Browse files
Files changed (1) hide show
  1. README.md +19 -44
README.md CHANGED
@@ -1,10 +1,6 @@
1
  ---
2
  tags:
3
- - automatic-speech-recognition
4
- - gary109/AI_Light_Dance
5
  - generated_from_trainer
6
- datasets:
7
- - ai_light_dance
8
  model-index:
9
  - name: ai-light-dance_singing_ft_pretrain_wav2vec2-large-lv60
10
  results: []
@@ -15,10 +11,10 @@ should probably proofread and complete it, then remove this comment. -->
15
 
16
  # ai-light-dance_singing_ft_pretrain_wav2vec2-large-lv60
17
 
18
- This model is a fine-tuned version of [gary109/ai-light-dance_pretrain_wav2vec2-large-lv60](https://huggingface.co/gary109/ai-light-dance_pretrain_wav2vec2-large-lv60) on the GARY109/AI_LIGHT_DANCE - ONSET-SINGING dataset.
19
  It achieves the following results on the evaluation set:
20
- - Loss: 1.5096
21
- - Wer: 0.9207
22
 
23
  ## Model description
24
 
@@ -38,51 +34,30 @@ More information needed
38
 
39
  The following hyperparameters were used during training:
40
  - learning_rate: 5e-05
41
- - train_batch_size: 12
42
- - eval_batch_size: 12
43
  - seed: 42
44
- - gradient_accumulation_steps: 8
45
- - total_train_batch_size: 96
46
  - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
47
- - lr_scheduler_type: linear
48
- - lr_scheduler_warmup_steps: 100
49
- - num_epochs: 30.0
50
  - mixed_precision_training: Native AMP
51
 
52
  ### Training results
53
 
54
  | Training Loss | Epoch | Step | Validation Loss | Wer |
55
  |:-------------:|:-----:|:----:|:---------------:|:------:|
56
- | 1.8237 | 1.0 | 184 | 1.9362 | 1.0743 |
57
- | 1.6396 | 2.0 | 368 | 1.7375 | 1.0047 |
58
- | 1.593 | 3.0 | 552 | 1.5942 | 0.9464 |
59
- | 1.5281 | 4.0 | 736 | 1.6037 | 0.9526 |
60
- | 1.5243 | 5.0 | 920 | 1.5292 | 0.9407 |
61
- | 1.4443 | 6.0 | 1104 | 1.5465 | 0.9457 |
62
- | 1.4801 | 7.0 | 1288 | 1.5965 | 0.9207 |
63
- | 1.4104 | 8.0 | 1472 | 1.5224 | 0.9211 |
64
- | 1.3828 | 9.0 | 1656 | 1.5400 | 0.9107 |
65
- | 1.3966 | 10.0 | 1840 | 1.5545 | 0.9289 |
66
- | 1.3481 | 11.0 | 2024 | 1.7372 | 0.9772 |
67
- | 1.3437 | 12.0 | 2208 | 1.5096 | 0.9207 |
68
- | 1.3207 | 13.0 | 2392 | 1.5559 | 0.9221 |
69
- | 1.3103 | 14.0 | 2576 | 1.5285 | 0.9242 |
70
- | 1.3032 | 15.0 | 2760 | 1.5297 | 0.8983 |
71
- | 1.2549 | 16.0 | 2944 | 1.5481 | 0.9075 |
72
- | 1.2484 | 17.0 | 3128 | 1.5886 | 0.9191 |
73
- | 1.2282 | 18.0 | 3312 | 1.6611 | 0.9231 |
74
- | 1.2168 | 19.0 | 3496 | 1.6368 | 0.9150 |
75
- | 1.2093 | 20.0 | 3680 | 1.7357 | 0.9135 |
76
- | 1.1951 | 21.0 | 3864 | 1.6973 | 0.9076 |
77
- | 1.1835 | 22.0 | 4048 | 1.7049 | 0.9170 |
78
- | 1.1554 | 23.0 | 4232 | 1.7663 | 0.9182 |
79
- | 1.1418 | 24.0 | 4416 | 1.6814 | 0.9134 |
80
- | 1.1391 | 25.0 | 4600 | 1.6740 | 0.8958 |
81
- | 1.1142 | 26.0 | 4784 | 1.7445 | 0.9015 |
82
- | 1.0598 | 27.0 | 4968 | 1.8278 | 0.9002 |
83
- | 1.0423 | 28.0 | 5152 | 1.8642 | 0.9042 |
84
- | 1.025 | 29.0 | 5336 | 1.9265 | 0.9078 |
85
- | 1.02 | 30.0 | 5520 | 1.9589 | 0.9056 |
86
 
87
 
88
  ### Framework versions
 
1
  ---
2
  tags:
 
 
3
  - generated_from_trainer
 
 
4
  model-index:
5
  - name: ai-light-dance_singing_ft_pretrain_wav2vec2-large-lv60
6
  results: []
 
11
 
12
  # ai-light-dance_singing_ft_pretrain_wav2vec2-large-lv60
13
 
14
+ This model is a fine-tuned version of [gary109/ai-light-dance_pretrain_wav2vec2-large-lv60](https://huggingface.co/gary109/ai-light-dance_pretrain_wav2vec2-large-lv60) on the None dataset.
15
  It achieves the following results on the evaluation set:
16
+ - Loss: 1.5296
17
+ - Wer: 0.9230
18
 
19
  ## Model description
20
 
 
34
 
35
  The following hyperparameters were used during training:
36
  - learning_rate: 5e-05
37
+ - train_batch_size: 2
38
+ - eval_batch_size: 2
39
  - seed: 42
40
+ - gradient_accumulation_steps: 16
41
+ - total_train_batch_size: 32
42
  - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
43
+ - lr_scheduler_type: cosine
44
+ - num_epochs: 10.0
 
45
  - mixed_precision_training: Native AMP
46
 
47
  ### Training results
48
 
49
  | Training Loss | Epoch | Step | Validation Loss | Wer |
50
  |:-------------:|:-----:|:----:|:---------------:|:------:|
51
+ | 1.6096 | 1.0 | 552 | 1.7650 | 1.0053 |
52
+ | 1.6294 | 2.0 | 1104 | 1.6735 | 0.9591 |
53
+ | 1.5509 | 3.0 | 1656 | 1.6170 | 0.9852 |
54
+ | 1.5175 | 4.0 | 2208 | 1.6312 | 0.9626 |
55
+ | 1.5267 | 5.0 | 2760 | 1.5032 | 0.9249 |
56
+ | 1.4055 | 6.0 | 3312 | 1.6107 | 0.9438 |
57
+ | 1.3267 | 7.0 | 3864 | 1.5386 | 0.9378 |
58
+ | 1.312 | 8.0 | 4416 | 1.4961 | 0.9206 |
59
+ | 1.3245 | 9.0 | 4968 | 1.5158 | 0.9182 |
60
+ | 1.2885 | 10.0 | 5520 | 1.5296 | 0.9230 |
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
61
 
62
 
63
  ### Framework versions