nguyenvulebinh
commited on
Commit
•
438b1f8
1
Parent(s):
a48472e
Create README.md
Browse files
README.md
ADDED
@@ -0,0 +1,52 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
language: ja
|
3 |
+
datasets:
|
4 |
+
- csj
|
5 |
+
tags:
|
6 |
+
- audio
|
7 |
+
- automatic-speech-recognition
|
8 |
+
license: cc-by-nc-4.0
|
9 |
+
---
|
10 |
+
|
11 |
+
### Usage
|
12 |
+
|
13 |
+
```python
|
14 |
+
#!pip install transformers==4.17.0
|
15 |
+
#!pip install https://github.com/kpu/kenlm/archive/master.zip
|
16 |
+
#!pip install pyctcdecode==0.4.0
|
17 |
+
from transformers.file_utils import cached_path, hf_bucket_url
|
18 |
+
from importlib.machinery import SourceFileLoader
|
19 |
+
from transformers import Wav2Vec2ProcessorWithLM
|
20 |
+
from IPython.lib.display import Audio
|
21 |
+
import torchaudio
|
22 |
+
import torch
|
23 |
+
|
24 |
+
# Load model & processor
|
25 |
+
model_name = "nguyenvulebinh/wav2vec2-base-ja"
|
26 |
+
model = SourceFileLoader("model", cached_path(hf_bucket_url(model_name,filename="model_handling.py"))).load_module().Wav2Vec2ForCTC.from_pretrained(model_name)
|
27 |
+
processor = Wav2Vec2ProcessorWithLM.from_pretrained(model_name)
|
28 |
+
|
29 |
+
# Load an example audio (16k)
|
30 |
+
audio, sample_rate = torchaudio.load(cached_path(hf_bucket_url(model_name, filename="sample.wav")))
|
31 |
+
input_data = processor.feature_extractor(audio[0], sampling_rate=16000, return_tensors='pt')
|
32 |
+
|
33 |
+
# Infer
|
34 |
+
output = model(**input_data)
|
35 |
+
|
36 |
+
# Output transcript without LM
|
37 |
+
print(processor.tokenizer.decode(output.logits.argmax(dim=-1)[0].detach().cpu().numpy()))
|
38 |
+
|
39 |
+
# Output transcript with LM
|
40 |
+
print(processor.decode(output.logits.cpu().detach().numpy()[0], beam_width=100).text)
|
41 |
+
```
|
42 |
+
|
43 |
+
### Model Parameters License
|
44 |
+
|
45 |
+
The ASR model parameters are made available for non-commercial use only, under the terms of the Creative Commons Attribution-NonCommercial 4.0 International (CC BY-NC 4.0) license. You can find details at: https://creativecommons.org/licenses/by-nc/4.0/legalcode
|
46 |
+
|
47 |
+
|
48 |
+
### Contact
|
49 |
+
|
50 |
+
nguyenvulebinh@gmail.com
|
51 |
+
|
52 |
+
[![Follow](https://img.shields.io/twitter/follow/nguyenvulebinh?style=social)](https://twitter.com/intent/follow?screen_name=nguyenvulebinh)
|