tensorops commited on
Commit
bab63a6
2 Parent(s): 57915bf d846ef3

Merge branch 'main' of https://huggingface.co/tensorops/whisper-medium-th-combined-v2 into main

Browse files
Files changed (1) hide show
  1. README.md +101 -0
README.md ADDED
@@ -0,0 +1,101 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ language:
3
+ - th
4
+ license: apache-2.0
5
+ tags:
6
+ - whisper-event
7
+ - generated_from_trainer
8
+ datasets:
9
+ - mozilla-foundation/common_voice_11_0
10
+ metrics:
11
+ - wer
12
+ model-index:
13
+ - name: Whisper Medium Thai Combined V2 - biodatlab
14
+ results:
15
+ - task:
16
+ name: Automatic Speech Recognition
17
+ type: automatic-speech-recognition
18
+ dataset:
19
+ name: mozilla-foundation/common_voice_11_0 th
20
+ type: mozilla-foundation/common_voice_11_0
21
+ config: th
22
+ split: test
23
+ args: th
24
+ metrics:
25
+ - name: Wer
26
+ type: wer
27
+ value: 13.03
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 Medium (Thai): Combined V2
34
+
35
+ This model is a fine-tuned version of [biodatlab/whisper-medium-th-1000iter](https://huggingface.co/biodatlab/whisper-medium-th-1000iter) on the mozilla-foundation/common_voice_11_0 th dataset.
36
+ It achieves the following results on the evaluation set:
37
+ - Loss: 0.1475
38
+ - WER: 13.03
39
+
40
+ ## Model description
41
+
42
+ Use the model with huggingface's `transformers` as follows:
43
+
44
+ ```py
45
+ from transformers import pipeline
46
+
47
+ MODEL_NAME = "biodatlab/whisper-medium-th-combined-v2" # specify the model name
48
+ lang = "th" # change to Thai langauge
49
+
50
+ device = 0 if torch.cuda.is_available() else "cpu"
51
+
52
+ pipe = pipeline(
53
+ task="automatic-speech-recognition",
54
+ model=MODEL_NAME,
55
+ chunk_length_s=30,
56
+ device=device,
57
+ )
58
+ pipe.model.config.forced_decoder_ids = pipe.tokenizer.get_decoder_prompt_ids(
59
+ language=lang,
60
+ task="transcribe"
61
+ )
62
+ text = pipe("audio.mp3")["text"] # give audio mp3 and transcribe text
63
+ ```
64
+
65
+
66
+ ## Intended uses & limitations
67
+
68
+ More information needed
69
+
70
+ ## Training and evaluation data
71
+
72
+ More information needed
73
+
74
+ ## Training procedure
75
+
76
+ ### Training hyperparameters
77
+
78
+ The following hyperparameters were used during training:
79
+ - learning_rate: 1e-05
80
+ - train_batch_size: 32
81
+ - eval_batch_size: 32
82
+ - seed: 42
83
+ - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
84
+ - lr_scheduler_type: linear
85
+ - lr_scheduler_warmup_steps: 500
86
+ - training_steps: 5000
87
+ - mixed_precision_training: Native AMP
88
+
89
+ ### Training results
90
+
91
+ | Training Loss | Epoch | Step | Validation Loss | Wer |
92
+ |:-------------:|:-----:|:----:|:---------------:|:-------:|
93
+ | 0.0679 | 2.09 | 5000 | 0.1475 | 13.03 |
94
+
95
+
96
+ ### Framework versions
97
+
98
+ - Transformers 4.26.0.dev0
99
+ - Pytorch 1.13.0
100
+ - Datasets 2.7.1
101
+ - Tokenizers 0.13.2