voxxer commited on
Commit
f5343c6
1 Parent(s): dad200c

update model card README.md

Browse files
Files changed (1) hide show
  1. README.md +98 -0
README.md ADDED
@@ -0,0 +1,98 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ language:
3
+ - en
4
+ license: apache-2.0
5
+ base_model: openai/whisper-tiny
6
+ tags:
7
+ - generated_from_trainer
8
+ datasets:
9
+ - PolyAI/minds14
10
+ metrics:
11
+ - wer
12
+ model-index:
13
+ - name: Whisper Tiny English
14
+ results:
15
+ - task:
16
+ name: Automatic Speech Recognition
17
+ type: automatic-speech-recognition
18
+ dataset:
19
+ name: Minds 14
20
+ type: PolyAI/minds14
21
+ config: en-US
22
+ split: train
23
+ args: en-US
24
+ metrics:
25
+ - name: Wer
26
+ type: wer
27
+ value: 0.258610624635143
28
+ ---
29
+
30
+ <!-- This model card has been generated automatically according to the information the Trainer had access to. You
31
+ should probably proofread and complete it, then remove this comment. -->
32
+
33
+ # Whisper Tiny English
34
+
35
+ This model is a fine-tuned version of [openai/whisper-tiny](https://huggingface.co/openai/whisper-tiny) on the Minds 14 dataset.
36
+ It achieves the following results on the evaluation set:
37
+ - Loss: 0.4154
38
+ - Wer Ortho: 0.2659
39
+ - Wer: 0.2586
40
+
41
+ ## Model description
42
+
43
+ More information needed
44
+
45
+ ## Intended uses & limitations
46
+
47
+ More information needed
48
+
49
+ ## Training and evaluation data
50
+
51
+ More information needed
52
+
53
+ ## Training procedure
54
+
55
+ ### Training hyperparameters
56
+
57
+ The following hyperparameters were used during training:
58
+ - learning_rate: 1e-05
59
+ - train_batch_size: 32
60
+ - eval_batch_size: 16
61
+ - seed: 42
62
+ - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
63
+ - lr_scheduler_type: constant_with_warmup
64
+ - lr_scheduler_warmup_steps: 20
65
+ - training_steps: 100
66
+
67
+ ### Training results
68
+
69
+ | Training Loss | Epoch | Step | Validation Loss | Wer Ortho | Wer |
70
+ |:-------------:|:-----:|:----:|:---------------:|:---------:|:------:|
71
+ | 4.2901 | 0.33 | 5 | 4.2556 | 0.4220 | 0.2919 |
72
+ | 4.3552 | 0.67 | 10 | 3.7784 | 0.4226 | 0.2931 |
73
+ | 3.453 | 1.0 | 15 | 2.9546 | 0.4152 | 0.2907 |
74
+ | 2.9147 | 1.33 | 20 | 2.4090 | 0.3988 | 0.2931 |
75
+ | 2.3042 | 1.67 | 25 | 1.7869 | 0.3701 | 0.3001 |
76
+ | 1.6056 | 2.0 | 30 | 1.1284 | 0.3494 | 0.3012 |
77
+ | 0.988 | 2.33 | 35 | 0.6892 | 0.3860 | 0.3403 |
78
+ | 0.6605 | 2.67 | 40 | 0.5611 | 0.3128 | 0.2849 |
79
+ | 0.4645 | 3.0 | 45 | 0.4982 | 0.3091 | 0.2901 |
80
+ | 0.4884 | 3.33 | 50 | 0.4640 | 0.2963 | 0.2855 |
81
+ | 0.404 | 3.67 | 55 | 0.4453 | 0.2884 | 0.2814 |
82
+ | 0.4745 | 4.0 | 60 | 0.4268 | 0.2762 | 0.2697 |
83
+ | 0.303 | 4.33 | 65 | 0.4182 | 0.2829 | 0.2720 |
84
+ | 0.2717 | 4.67 | 70 | 0.4119 | 0.2829 | 0.2750 |
85
+ | 0.3464 | 5.0 | 75 | 0.4080 | 0.2860 | 0.2761 |
86
+ | 0.2193 | 5.33 | 80 | 0.4054 | 0.2823 | 0.2750 |
87
+ | 0.2138 | 5.67 | 85 | 0.4064 | 0.2762 | 0.2680 |
88
+ | 0.1571 | 6.0 | 90 | 0.4102 | 0.2799 | 0.2715 |
89
+ | 0.1398 | 6.33 | 95 | 0.4146 | 0.2768 | 0.2697 |
90
+ | 0.1523 | 6.67 | 100 | 0.4154 | 0.2659 | 0.2586 |
91
+
92
+
93
+ ### Framework versions
94
+
95
+ - Transformers 4.31.0
96
+ - Pytorch 2.0.1+cu118
97
+ - Datasets 2.14.4
98
+ - Tokenizers 0.13.3