Hugging Face's logo

DahmL
/
wav2vec2-large-xls-r-300m-turkish-colab

Automatic Speech Recognition
Transformers
TensorBoard
Safetensors
wav2vec2
Generated from Trainer
Eval Results
Inference Endpoints
Model card Files Metrics Community
DahmL commited on
Commit
db82436
•
1 Parent(s): 18bb174

End of training

Browse files
Files changed (2) hide show
  1. README.md +16 -13
  2. model.safetensors +1 -1
README.md CHANGED
@@ -22,7 +22,7 @@ model-index:
22
  metrics:
23
  - name: Wer
24
  type: wer
25
- value: 0.87
26
  ---
27
 
28
  <!-- This model card has been generated automatically according to the information the Trainer had access to. You
@@ -32,8 +32,8 @@ should probably proofread and complete it, then remove this comment. -->
32
 
33
  This model is a fine-tuned version of [facebook/wav2vec2-xls-r-300m](https://huggingface.co/facebook/wav2vec2-xls-r-300m) on the common_voice_13_0 dataset.
34
  It achieves the following results on the evaluation set:
35
- - Loss: 0.8038
36
- - Wer: 0.87
37
 
38
  ## Model description
39
 
@@ -53,28 +53,31 @@ More information needed
53
 
54
  The following hyperparameters were used during training:
55
  - learning_rate: 0.0003
56
- - train_batch_size: 16
57
- - eval_batch_size: 8
58
  - seed: 42
59
  - gradient_accumulation_steps: 2
60
- - total_train_batch_size: 32
61
  - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
62
  - lr_scheduler_type: linear
63
- - lr_scheduler_warmup_steps: 500
64
- - num_epochs: 30
65
  - mixed_precision_training: Native AMP
66
 
67
  ### Training results
68
 
69
- | Training Loss | Epoch | Step | Validation Loss | Wer |
70
- |:-------------:|:-----:|:----:|:---------------:|:-----:|
71
- | 6.0471 | 12.7 | 400 | 0.9101 | 0.941 |
72
- | 0.3753 | 25.4 | 800 | 0.8038 | 0.87 |
 
 
 
73
 
74
 
75
  ### Framework versions
76
 
77
  - Transformers 4.36.1
78
- - Pytorch 2.1.0+cu121
79
  - Datasets 2.15.0
80
  - Tokenizers 0.15.0
 
22
  metrics:
23
  - name: Wer
24
  type: wer
25
+ value: 0.7091714338438826
26
  ---
27
 
28
  <!-- This model card has been generated automatically according to the information the Trainer had access to. You
 
32
 
33
  This model is a fine-tuned version of [facebook/wav2vec2-xls-r-300m](https://huggingface.co/facebook/wav2vec2-xls-r-300m) on the common_voice_13_0 dataset.
34
  It achieves the following results on the evaluation set:
35
+ - Loss: 0.5247
36
+ - Wer: 0.7092
37
 
38
  ## Model description
39
 
 
53
 
54
  The following hyperparameters were used during training:
55
  - learning_rate: 0.0003
56
+ - train_batch_size: 64
57
+ - eval_batch_size: 16
58
  - seed: 42
59
  - gradient_accumulation_steps: 2
60
+ - total_train_batch_size: 128
61
  - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
62
  - lr_scheduler_type: linear
63
+ - lr_scheduler_warmup_steps: 300
64
+ - num_epochs: 20
65
  - mixed_precision_training: Native AMP
66
 
67
  ### Training results
68
 
69
+ | Training Loss | Epoch | Step | Validation Loss | Wer |
70
+ |:-------------:|:-----:|:----:|:---------------:|:------:|
71
+ | 0.1953 | 3.86 | 400 | 0.5740 | 0.7963 |
72
+ | 0.1959 | 7.73 | 800 | 0.5169 | 0.7743 |
73
+ | 0.1486 | 11.59 | 1200 | 0.5334 | 0.7501 |
74
+ | 0.1146 | 15.46 | 1600 | 0.5186 | 0.7226 |
75
+ | 0.0885 | 19.32 | 2000 | 0.5247 | 0.7092 |
76
 
77
 
78
  ### Framework versions
79
 
80
  - Transformers 4.36.1
81
+ - Pytorch 1.10.0+cu113
82
  - Datasets 2.15.0
83
  - Tokenizers 0.15.0
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:2b6e92c19e8fd579ac9a9c70825f33963fb02143903b883b6f043561ffe3f347
3
  size 1261967332
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:01b71a384a0ebe3caa4db77c198ea080996f4719e4d0bfe63721fa65340c5263
3
  size 1261967332