Speech2text

What is it?

This project includes a wrapper around speech-to-text models from the Transformers library with a graphical interface provided by Gradio. Speech2Text is a modular project that can be integrated with other projects requiring a speech-to-text model. The server-side implementation is built on FastAPI. If needed, the project can be used independently by accessing the root page of the deployed project.

Table of contents

speech2text_factory/
│
├── docker-compose.yml              # Docker Compose configuration file
├── dockerfile                      # Dockerfile for creating Docker container
├── LICENSE                         # License file
├── Makefile                        # Makefile for project automation
├── README.md                       # Project description and instructions
├── requirements.txt                # Python dependencies list for pip
├── docs/                           # Directory for documentation and design files
│   ├── index.html                  # Main HTML documentation file
│   ├── src/                        # Directory for source files used in documentation
│   └── style.css                   # CSS stylesheet for documentation styling
│
└── src/                            # Source code directory
    ├── ai_models/                  # Directory for AI model scripts
    │   ├── weights/                # Directory for model weights
    │   └── model.py                # Speech-to-text model script
    │
    ├── api/                        # Directory for API scripts
    │   ├── app/                    # Directory for application specific API scripts
    │   │   ├── crud.py             # Script for CRUD operations
    │   │   ├── endpoint.py         # Endpoint script
    │   │   ├── middleware.py       # Middleware script
    │   │   ├── models.py           # Models script
    │   └── main.py                 # Main API script
    │    
    ├── main.py                     # Main script
    └── utils/                      # Directory for utility scripts
        └── features_extractor.py   # Feature extractor script

Installation

Locally

To install all dependencies and run the project locally, follow these steps:

1. Build the project:

   make build

2. Navigate to the src directory:

   cd src

3. Run the main script from src directory:

   python3 main.py

Additionally, create a .env file with the variables MODEL_NAME and USE_CUDA. By default, these are set to openai/whisper-small and True respectively.

Docker Installation

To build and run a Docker container with a specific Speech2Text model, follow these steps:

1. Build the Docker image: In the speech2text directory, run:

   docker build -t speech2text .

   This command creates a Docker image named speech2text.

2. Run the Docker container:

   docker run -p PORT:8000 --gpus all --name CONTAINER_NAME --env MODEL_NAME="model-name" --env USE_CUDA="True" speech2text

   Replace PORT with the port number on which you intend to run the container and model-name with the desired speech-to-text model from Hugging Face.

Explanation:

• -p PORT:8000: Maps PORT on your host to port 8000 in the container.
• --gpus all: Allocates all available GPUs to the container.
• --name CONTAINER_NAME: Assigns a name to the container.
• --env MODEL_NAME="model-name": Sets the environment variable MODEL_NAME in the container.
• --env USE_CUDA="True": Sets the environment variable USE_CUDA in the container.

To use CUDA in container you have to install nvidia-docekr2. Just run this comand in cmd

sudo apt-get install -y nvidia-docker2 

API Endpoints

Endpoint	Description
/	Return a greeting site.
/predict	Return a result of speech2text model.
/get-config	Return the config of the model.

Contribution

Feel free to open an issue or submit a pull request.