Hugging Face's logo

HeinrichWirth
/
whisper-tiny_en

Automatic Speech Recognition
Transformers
PyTorch
whisper
Generated from Trainer
Eval Results
Inference Endpoints
Model card Files Community
HeinrichWirth commited on
Commit
b5ee566
•
1 Parent(s): 1a1d820

End of training

Browse files
Files changed (3) hide show
  1. README.md +8 -43
  2. pytorch_model.bin +1 -1
  3. training_args.bin +1 -1
README.md CHANGED
@@ -32,8 +32,8 @@ should probably proofread and complete it, then remove this comment. -->
32
 
33
  This model is a fine-tuned version of [openai/whisper-tiny](https://huggingface.co/openai/whisper-tiny) on the PolyAI/minds14 dataset.
34
  It achieves the following results on the evaluation set:
35
- - Loss: 0.5991
36
- - Wer Ortho: 36.0700
37
  - Wer: 34.5607
38
 
39
  ## Model description
@@ -60,52 +60,17 @@ The following hyperparameters were used during training:
60
  - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
61
  - lr_scheduler_type: constant_with_warmup
62
  - lr_scheduler_warmup_steps: 50
63
- - training_steps: 1000
64
 
65
  ### Training results
66
 
67
  | Training Loss | Epoch | Step | Validation Loss | Wer Ortho | Wer |
68
  |:-------------:|:-----:|:----:|:---------------:|:---------:|:-------:|
69
- | 4.3106 | 0.89 | 25 | 4.5196 | 51.0094 | 36.6925 |
70
- | 3.9031 | 1.79 | 50 | 3.5872 | 51.0094 | 37.0155 |
71
- | 2.9832 | 2.68 | 75 | 2.8616 | 48.9906 | 36.7571 |
72
- | 2.3893 | 3.57 | 100 | 2.3230 | 47.1063 | 37.0155 |
73
- | 1.9074 | 4.46 | 125 | 1.8030 | 44.6837 | 36.8863 |
74
- | 1.3515 | 5.36 | 150 | 1.3467 | 43.5397 | 37.1447 |
75
- | 0.9984 | 6.25 | 175 | 0.9622 | 42.6649 | 37.0155 |
76
- | 0.7155 | 7.14 | 200 | 0.7915 | 42.1265 | 37.1447 |
77
- | 0.6152 | 8.04 | 225 | 0.7200 | 41.1844 | 37.1447 |
78
- | 0.523 | 8.93 | 250 | 0.6814 | 40.9152 | 37.1447 |
79
- | 0.4744 | 9.82 | 275 | 0.6561 | 39.2328 | 35.8527 |
80
- | 0.4411 | 10.71 | 300 | 0.6371 | 40.4441 | 37.4031 |
81
- | 0.3926 | 11.61 | 325 | 0.6194 | 37.5505 | 35.2067 |
82
- | 0.3855 | 12.5 | 350 | 0.6111 | 37.2140 | 35.0775 |
83
- | 0.3709 | 13.39 | 375 | 0.6012 | 37.3486 | 35.1421 |
84
- | 0.3315 | 14.29 | 400 | 0.5963 | 37.3486 | 35.2713 |
85
- | 0.3113 | 15.18 | 425 | 0.5892 | 37.2813 | 35.4005 |
86
- | 0.3112 | 16.07 | 450 | 0.5849 | 37.3486 | 35.4005 |
87
- | 0.2809 | 16.96 | 475 | 0.5827 | 36.6756 | 34.7545 |
88
- | 0.2694 | 17.86 | 500 | 0.5778 | 36.8775 | 34.9483 |
89
- | 0.236 | 18.75 | 525 | 0.5775 | 36.2046 | 34.2377 |
90
- | 0.2512 | 19.64 | 550 | 0.5755 | 36.6756 | 34.6899 |
91
- | 0.2154 | 20.54 | 575 | 0.5760 | 36.2046 | 34.3023 |
92
- | 0.2174 | 21.43 | 600 | 0.5742 | 36.0027 | 34.1085 |
93
- | 0.1923 | 22.32 | 625 | 0.5741 | 36.2046 | 34.3669 |
94
- | 0.198 | 23.21 | 650 | 0.5731 | 36.1373 | 34.3669 |
95
- | 0.1699 | 24.11 | 675 | 0.5757 | 36.2719 | 34.5607 |
96
- | 0.1662 | 25.0 | 700 | 0.5768 | 36.6083 | 34.9483 |
97
- | 0.1559 | 25.89 | 725 | 0.5762 | 36.8775 | 35.1421 |
98
- | 0.141 | 26.79 | 750 | 0.5801 | 36.9448 | 35.2713 |
99
- | 0.1318 | 27.68 | 775 | 0.5791 | 36.8775 | 35.2067 |
100
- | 0.1342 | 28.57 | 800 | 0.5809 | 37.0794 | 35.2713 |
101
- | 0.1179 | 29.46 | 825 | 0.5829 | 36.8775 | 35.0775 |
102
- | 0.1111 | 30.36 | 850 | 0.5835 | 37.2140 | 35.4005 |
103
- | 0.0993 | 31.25 | 875 | 0.5887 | 37.0794 | 35.4005 |
104
- | 0.0956 | 32.14 | 900 | 0.5900 | 36.6756 | 35.0129 |
105
- | 0.0886 | 33.04 | 925 | 0.5944 | 36.7429 | 35.0775 |
106
- | 0.0811 | 33.93 | 950 | 0.5945 | 36.3392 | 34.6899 |
107
- | 0.0748 | 34.82 | 975 | 0.6010 | 36.6083 | 35.0775 |
108
- | 0.0662 | 35.71 | 1000 | 0.5991 | 36.0700 | 34.5607 |
109
 
110
 
111
  ### Framework versions
 
32
 
33
  This model is a fine-tuned version of [openai/whisper-tiny](https://huggingface.co/openai/whisper-tiny) on the PolyAI/minds14 dataset.
34
  It achieves the following results on the evaluation set:
35
+ - Loss: 0.5771
36
+ - Wer Ortho: 0.3668
37
  - Wer: 34.5607
38
 
39
  ## Model description
 
60
  - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
61
  - lr_scheduler_type: constant_with_warmup
62
  - lr_scheduler_warmup_steps: 50
63
+ - training_steps: 500
64
 
65
  ### Training results
66
 
67
  | Training Loss | Epoch | Step | Validation Loss | Wer Ortho | Wer |
68
  |:-------------:|:-----:|:----:|:---------------:|:---------:|:-------:|
69
+ | 2.3696 | 3.57 | 100 | 2.3300 | 0.4744 | 36.8863 |
70
+ | 0.7034 | 7.14 | 200 | 0.7943 | 0.4206 | 37.0801 |
71
+ | 0.466 | 10.71 | 300 | 0.6352 | 0.3923 | 36.1111 |
72
+ | 0.3367 | 14.29 | 400 | 0.5936 | 0.3863 | 36.4987 |
73
+ | 0.2717 | 17.86 | 500 | 0.5771 | 0.3668 | 34.5607 |
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
74
 
75
 
76
  ### Framework versions
pytorch_model.bin CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:9fe101d3e88e0f926802e2ac7bcaf5c851ad1ad869a5465ac940875e5c24a085
3
  size 151099049
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:40fbf97e7dd6d3471c3a4615a401986daecc92bb24f9e15a2874c2b8e9710ef5
3
  size 151099049
training_args.bin CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:510126af2d7b1b2b8dc033f57195b10b33b4c3351ec342e99524badf629e5799
3
  size 4219
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:aec44140b675f665bb52c18f1438438db8cf485a3092bb84f843c571f5365df5
3
  size 4219