seastar105 commited on
Commit
faedbf5
1 Parent(s): bbafd4d

update model card README.md

Browse files
Files changed (1) hide show
  1. README.md +87 -0
README.md ADDED
@@ -0,0 +1,87 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ language:
3
+ - ko
4
+ license: apache-2.0
5
+ tags:
6
+ - hf-asr-leaderboard
7
+ - whisper-event
8
+ - generated_from_trainer
9
+ datasets:
10
+ - kresnik/zeroth_korean
11
+ metrics:
12
+ - wer
13
+ model-index:
14
+ - name: Whisper Medium Korean
15
+ results:
16
+ - task:
17
+ name: Automatic Speech Recognition
18
+ type: automatic-speech-recognition
19
+ dataset:
20
+ name: Zeroth Korean
21
+ type: kresnik/zeroth_korean
22
+ config: clean
23
+ split: test
24
+ args: 'split: test'
25
+ metrics:
26
+ - name: Wer
27
+ type: wer
28
+ value: 3.6440295136274656
29
+ ---
30
+
31
+ <!-- This model card has been generated automatically according to the information the Trainer had access to. You
32
+ should probably proofread and complete it, then remove this comment. -->
33
+
34
+ # Whisper Medium Korean
35
+
36
+ This model is a fine-tuned version of [openai/whisper-medium](https://huggingface.co/openai/whisper-medium) on the Zeroth Korean dataset.
37
+ It achieves the following results on the evaluation set:
38
+ - Loss: 0.0727
39
+ - Wer: 3.6440
40
+ - Cer: 1.4840
41
+
42
+ ## Model description
43
+
44
+ More information needed
45
+
46
+ ## Intended uses & limitations
47
+
48
+ More information needed
49
+
50
+ ## Training and evaluation data
51
+
52
+ More information needed
53
+
54
+ ## Training procedure
55
+
56
+ ### Training hyperparameters
57
+
58
+ The following hyperparameters were used during training:
59
+ - learning_rate: 5e-06
60
+ - train_batch_size: 8
61
+ - eval_batch_size: 8
62
+ - seed: 42
63
+ - gradient_accumulation_steps: 2
64
+ - total_train_batch_size: 16
65
+ - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
66
+ - lr_scheduler_type: linear
67
+ - lr_scheduler_warmup_steps: 500
68
+ - training_steps: 5000
69
+ - mixed_precision_training: Native AMP
70
+
71
+ ### Training results
72
+
73
+ | Training Loss | Epoch | Step | Validation Loss | Wer | Cer |
74
+ |:-------------:|:-----:|:----:|:---------------:|:------:|:------:|
75
+ | 0.0873 | 0.72 | 1000 | 0.1086 | 7.7549 | 2.5597 |
76
+ | 0.0258 | 1.44 | 2000 | 0.0805 | 4.5475 | 1.7588 |
77
+ | 0.0091 | 2.16 | 3000 | 0.0719 | 3.7946 | 1.5664 |
78
+ | 0.0086 | 2.88 | 4000 | 0.0704 | 3.5537 | 1.5232 |
79
+ | 0.0019 | 3.59 | 5000 | 0.0727 | 3.6440 | 1.4840 |
80
+
81
+
82
+ ### Framework versions
83
+
84
+ - Transformers 4.26.0.dev0
85
+ - Pytorch 1.13.0a0+d0d6b1f
86
+ - Datasets 2.7.1
87
+ - Tokenizers 0.13.2