Wav2vec NCKH 2022

🚀 Wav2vec2 NCKH Vietnamese 2022

This is a speech recognition model based on the Transformer architecture, which can perform automatic speech recognition tasks on Vietnamese datasets.

🚀 Quick Start

Convert from model .pt to transformer

You can convert the model from .pt format to the Transformer format by following these steps:

pip install transformers[sentencepiece]
pip install fairseq -U
git clone https://github.com/huggingface/transformers.git
cp transformers/src/transformers/models/wav2vec2/convert_wav2vec2_original_pytorch_checkpoint_to_pytorch.py .
wget https://dl.fbaipublicfiles.com/fairseq/wav2vec/wav2vec_small.pt -O ./wav2vec_small.pt
mkdir dict
wget https://dl.fbaipublicfiles.com/fairseq/wav2vec/dict.ltr.txt
mkdir outputs
python convert_wav2vec2_original_pytorch_checkpoint_to_pytorch.py 
--pytorch_dump_folder_path ./outputs --checkpoint_path ./finetuned/wav2vec_small.pt
 --dict_path ./dict/dict.ltr.txt --not_finetuned

Install and upload model

To install and upload the model, you can use the following commands:

curl -s https://packagecloud.io/install/repositories/github/git-lfs/script.deb.sh | sudo bash
git lfs install
sudo apt-get install git-lfs
git lfs install
git clone https://huggingface.co/hoangbinhmta99/wav2vec-demo
ls
cd wav2vec-demo/
git status
git add .
git commit -m "First model version"
git config --global user.email [yourname]
git config --global user.name [yourpass]
git commit -m "First model version"
git push

✨ Features

• Datasets: This model is trained on the vivos and common_voice datasets.
• Metrics: The model uses the wer (Word Error Rate) metric for evaluation.
• Pipeline Tag: It belongs to the automatic-speech-recognition pipeline.

📦 Installation

The installation steps are included in the "Quick Start" section. You can follow the commands to install the necessary dependencies and clone the model repository.

📚 Documentation

Model Information

Model Results

• Task: Speech Recognition (automatic-speech-recognition)
• Dataset: Common Voice vi (common_voice with args vi)
• Metrics:
  • Test WER: No

📄 License

This model is licensed under the cc-by-nc-4.0 license.