wav2vec2-large-vietnamese-vivos

Wav2Vec2-Large-Vietnamese-VIVOS (Fine-tuned)

This repository contains a fine-tuned version of the nguyenvulebinh/wav2vec2-large-vi-vlsp2020 model on the VIVOS: Vietnamese Speech Corpus for ASR.
The model is designed for Automatic Speech Recognition (ASR) in Vietnamese.

---

📖 Dataset: VIVOS

• Name: VIVOS - Vietnamese Speech Corpus for ASR
• Size: ~15 hours of recording speech
• Description:
  VIVOS is a free Vietnamese speech corpus prepared for Automatic Speech Recognition (ASR) tasks.
  It consists of audio recordings paired with corresponding transcriptions.
• More information: Kaggle dataset link

---

🧩 Base Model

This model is fine-tuned from nguyenvulebinh/wav2vec2-large-vi-vlsp2020,
which is a large-scale pretrained Wav2Vec2 model for Vietnamese ASR.

---

🚀 Usage

You can use this model directly with the Hugging Face transformers library:

from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
import torch
import soundfile as sf

# Load processor and model
processor = Wav2Vec2Processor.from_pretrained("your-username/your-model-repo")
model = Wav2Vec2ForCTC.from_pretrained("your-username/your-model-repo")

# Load audio file
speech, rate = sf.read("path_to_your_audio.wav")

# Preprocess
input_values = processor(speech, sampling_rate=rate, return_tensors="pt", padding=True).input_values

# Perform inference
with torch.no_grad():
    logits = model(input_values).logits

# Decode prediction
predicted_ids = torch.argmax(logits, dim=-1)
transcription = processor.decode(predicted_ids[0])

print("Transcription:", transcription)

📌 Acknowledgements

• Base model: nguyenvulebinh/wav2vec2-large-vi-vlsp2020
• Dataset: VIVOS - Vietnamese Speech Corpus for ASR