Hugging Face's logo

jerrychatz
/
wav2vec2-large-xls-r-300m-greek

Automatic Speech Recognition
Transformers
PyTorch
TensorBoard
wav2vec2
Generated from Trainer
Inference Endpoints
Model card Files Metrics Community
jerrychatz commited on
Commit
6c41361
·
1 Parent(s): 9b87d06

update model card README.md

Browse files
Files changed (1) hide show
  1. README.md +20 -21
README.md CHANGED
@@ -15,8 +15,8 @@ should probably proofread and complete it, then remove this comment. -->
15
 
16
  This model was trained from scratch on the common_voice dataset.
17
  It achieves the following results on the evaluation set:
18
- - Loss: 0.4860
19
- - Wer: 0.3464
20
 
21
  ## Model description
22
 
@@ -35,7 +35,7 @@ More information needed
35
  ### Training hyperparameters
36
 
37
  The following hyperparameters were used during training:
38
- - learning_rate: 0.0003
39
  - train_batch_size: 16
40
  - eval_batch_size: 8
41
  - seed: 42
@@ -43,7 +43,6 @@ The following hyperparameters were used during training:
43
  - total_train_batch_size: 32
44
  - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
45
  - lr_scheduler_type: linear
46
- - lr_scheduler_warmup_steps: 500
47
  - num_epochs: 30
48
  - mixed_precision_training: Native AMP
49
 
@@ -51,23 +50,23 @@ The following hyperparameters were used during training:
51
 
52
  | Training Loss | Epoch | Step | Validation Loss | Wer |
53
  |:-------------:|:-----:|:----:|:---------------:|:------:|
54
- | 3.1726 | 1.72 | 200 | 0.6130 | 0.4181 |
55
- | 0.1165 | 3.45 | 400 | 0.6819 | 0.4588 |
56
- | 0.1272 | 5.17 | 600 | 0.5812 | 0.4385 |
57
- | 0.1253 | 6.89 | 800 | 0.5866 | 0.4723 |
58
- | 0.1136 | 8.62 | 1000 | 0.5289 | 0.4361 |
59
- | 0.0936 | 10.34 | 1200 | 0.5785 | 0.4265 |
60
- | 0.0853 | 12.07 | 1400 | 0.5314 | 0.4365 |
61
- | 0.0761 | 13.79 | 1600 | 0.5065 | 0.4056 |
62
- | 0.0707 | 15.52 | 1800 | 0.5427 | 0.3948 |
63
- | 0.0632 | 17.24 | 2000 | 0.5227 | 0.3840 |
64
- | 0.0501 | 18.96 | 2200 | 0.5150 | 0.3875 |
65
- | 0.0462 | 20.69 | 2400 | 0.4949 | 0.3747 |
66
- | 0.0402 | 22.41 | 2600 | 0.4773 | 0.3720 |
67
- | 0.0339 | 24.14 | 2800 | 0.4719 | 0.3593 |
68
- | 0.0302 | 25.86 | 3000 | 0.4995 | 0.3540 |
69
- | 0.028 | 27.58 | 3200 | 0.4840 | 0.3510 |
70
- | 0.0244 | 29.31 | 3400 | 0.4860 | 0.3464 |
71
 
72
 
73
  ### Framework versions
 
15
 
16
  This model was trained from scratch on the common_voice dataset.
17
  It achieves the following results on the evaluation set:
18
+ - Loss: 0.4823
19
+ - Wer: 0.3338
20
 
21
  ## Model description
22
 
 
35
  ### Training hyperparameters
36
 
37
  The following hyperparameters were used during training:
38
+ - learning_rate: 3e-05
39
  - train_batch_size: 16
40
  - eval_batch_size: 8
41
  - seed: 42
 
43
  - total_train_batch_size: 32
44
  - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
45
  - lr_scheduler_type: linear
 
46
  - num_epochs: 30
47
  - mixed_precision_training: Native AMP
48
 
 
50
 
51
  | Training Loss | Epoch | Step | Validation Loss | Wer |
52
  |:-------------:|:-----:|:----:|:---------------:|:------:|
53
+ | 0.0106 | 1.72 | 200 | 0.5519 | 0.3537 |
54
+ | 0.0249 | 3.45 | 400 | 0.5174 | 0.3465 |
55
+ | 0.0206 | 5.17 | 600 | 0.4721 | 0.3323 |
56
+ | 0.0221 | 6.89 | 800 | 0.4652 | 0.3373 |
57
+ | 0.0204 | 8.62 | 1000 | 0.4883 | 0.3389 |
58
+ | 0.0192 | 10.34 | 1200 | 0.4785 | 0.3389 |
59
+ | 0.0186 | 12.07 | 1400 | 0.4789 | 0.3378 |
60
+ | 0.0172 | 13.79 | 1600 | 0.4915 | 0.3347 |
61
+ | 0.0184 | 15.52 | 1800 | 0.4759 | 0.3440 |
62
+ | 0.0168 | 17.24 | 2000 | 0.4891 | 0.3371 |
63
+ | 0.0155 | 18.96 | 2200 | 0.4928 | 0.3394 |
64
+ | 0.0146 | 20.69 | 2400 | 0.4834 | 0.3357 |
65
+ | 0.0146 | 22.41 | 2600 | 0.4814 | 0.3362 |
66
+ | 0.0151 | 24.14 | 2800 | 0.4791 | 0.3345 |
67
+ | 0.0136 | 25.86 | 3000 | 0.4825 | 0.3356 |
68
+ | 0.0136 | 27.58 | 3200 | 0.4850 | 0.3351 |
69
+ | 0.0127 | 29.31 | 3400 | 0.4823 | 0.3338 |
70
 
71
 
72
  ### Framework versions