File size: 897 Bytes
3775c9d 55e9761 bbf4396 c7729c4 bbf4396 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 |
---
license: mit
language:
- vi
metrics:
- wer
base_model:
- facebook/wav2vec2-xls-r-300m
pipeline_tag: automatic-speech-recognition
---
Tôi đã fine-tune với 15Gb dữ liệu audio với kết quả Wer: 24.46
## Cách sử dụng
```
import torch
from transformers import Wav2Vec2Processor, Wav2Vec2ForCTC
import torchaudio
mydevice = 'cuda'
processor = Wav2Vec2Processor.from_pretrained("hataphu/wav2vec2-vi-300m")
model = Wav2Vec2ForCTC.from_pretrained("hataphu/wav2vec2-vi-300m")
model.to(mydevice)
model.eval()
audio_input, sampling_rate = torchaudio.load('audio-path-file')
input_values = processor(
audio_input.squeeze().numpy(), sampling_rate=sampling_rate
).input_values[0]
logits = model(torch.tensor(input_values).unsqueeze(0).to(mydevice)).logits
predicted_ids = torch.argmax(logits, dim=-1)
transcription = processor.decode(predicted_ids[0])
print(transcription)
```
|