Hugging Face's logo

kawther1
/
whisper-LARGE-AR

PEFT
TensorBoard
Safetensors
Generated from Trainer
Model card Files Metrics Community
kawther1 commited on
Commit
f6038ee
·
verified ·
1 Parent(s): 99e939d

END

Browse files
Files changed (1) hide show
  1. README.md +16 -17
README.md CHANGED
@@ -3,9 +3,9 @@ license: apache-2.0
3
  library_name: peft
4
  tags:
5
  - generated_from_trainer
6
- base_model: openai/whisper-large
7
  datasets:
8
- - common_voice_11_0
9
  metrics:
10
  - wer
11
  model-index:
@@ -18,11 +18,11 @@ should probably proofread and complete it, then remove this comment. -->
18
 
19
  # whisper-LARGE-AR
20
 
21
- This model is a fine-tuned version of [openai/whisper-large](https://huggingface.co/openai/whisper-large) on the common_voice_11_0 dataset.
22
  It achieves the following results on the evaluation set:
23
- - Loss: 2.0206
24
- - Wer Ortho: 53.5644
25
- - Wer: 56.1417
26
 
27
  ## Model description
28
 
@@ -41,22 +41,21 @@ More information needed
41
  ### Training hyperparameters
42
 
43
  The following hyperparameters were used during training:
44
- - learning_rate: 0.0002
45
- - train_batch_size: 4
46
- - eval_batch_size: 4
47
  - seed: 42
48
- - gradient_accumulation_steps: 4
49
- - total_train_batch_size: 16
50
  - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
51
  - lr_scheduler_type: constant_with_warmup
52
- - lr_scheduler_warmup_steps: 10
53
- - training_steps: 10
 
54
 
55
  ### Training results
56
 
57
- | Training Loss | Epoch | Step | Validation Loss | Wer Ortho | Wer |
58
- |:-------------:|:-----:|:----:|:---------------:|:---------:|:-------:|
59
- | 1.3305 | 0.16 | 10 | 2.0206 | 53.5644 | 56.1417 |
60
 
61
 
62
  ### Framework versions
@@ -64,5 +63,5 @@ The following hyperparameters were used during training:
64
  - PEFT 0.11.2.dev0
65
  - Transformers 4.42.0.dev0
66
  - Pytorch 2.3.0+cu121
67
- - Datasets 2.19.1
68
  - Tokenizers 0.19.1
 
3
  library_name: peft
4
  tags:
5
  - generated_from_trainer
6
+ base_model: openai/whisper-large-v2
7
  datasets:
8
+ - common_voice_16_1
9
  metrics:
10
  - wer
11
  model-index:
 
18
 
19
  # whisper-LARGE-AR
20
 
21
+ This model is a fine-tuned version of [openai/whisper-large-v2](https://huggingface.co/openai/whisper-large-v2) on the common_voice_16_1 dataset.
22
  It achieves the following results on the evaluation set:
23
+ - Loss: 2.5181
24
+ - Wer Ortho: 47.6821
25
+ - Wer: 48.9362
26
 
27
  ## Model description
28
 
 
41
  ### Training hyperparameters
42
 
43
  The following hyperparameters were used during training:
44
+ - learning_rate: 5e-06
45
+ - train_batch_size: 16
46
+ - eval_batch_size: 16
47
  - seed: 42
 
 
48
  - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
49
  - lr_scheduler_type: constant_with_warmup
50
+ - lr_scheduler_warmup_steps: 20
51
+ - training_steps: 100
52
+ - mixed_precision_training: Native AMP
53
 
54
  ### Training results
55
 
56
+ | Training Loss | Epoch | Step | Validation Loss | Wer Ortho | Wer |
57
+ |:-------------:|:-------:|:----:|:---------------:|:---------:|:-------:|
58
+ | 1.6571 | 14.2857 | 100 | 2.5181 | 47.6821 | 48.9362 |
59
 
60
 
61
  ### Framework versions
 
63
  - PEFT 0.11.2.dev0
64
  - Transformers 4.42.0.dev0
65
  - Pytorch 2.3.0+cu121
66
+ - Datasets 2.19.2
67
  - Tokenizers 0.19.1