Downloads · 30 days
12
6% of all-time downloads
wav2vec2-vi-300m
hataphu/wav2vec2-vi-300m
wav2vec2-vi-300m is a automatic speech recognition model from hataphu. Use it when you need speech turned into text. The card lists the license as mit.
Tôi đã fine-tune với 15Gb dữ liệu audio với kết quả Wer: 24.46
All-time downloads
195
Public
Parameters
316M
1.3 GB on disk
Likes
1
Public
Repo makeup
Click a slice to open those files.
.safetensors1.3 GB · 100%
At a glance
- Task
- Automatic Speech Recognition
- License
- mit
- Model type
- wav2vec2
- Access
- Public
- Created
- Nov 11, 2024
- Updated
- Nov 12, 2024
- SHA
- 93592715
Base models
Model card
From the Hugging Face model README
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)
- Task
- Automatic Speech Recognition
- Type
- wav2vec2
- License
- mit
- Languages
- vi
- Created
- Nov 11, 2024
- Updated
- Nov 12, 2024