ahmedewis / ml_deployment

Create FastAPI app for deploying the HiFi GAN model

Deploying AI Models on Cloud

The repo contains code for deploying a solution on AWS. We will be using Fast API for building the app.

Environment

• fastapi==0.63.0
• librosa==0.8.0
• matplotlib==3.1.2
• numpy==1.19.5
• Pillow==8.1.0
• pypesq==1.2.4
• pystoi==0.3.3
• starlette==0.13.6
• torch==1.7.1
• torchaudio==0.7.2
• tqdm==4.47.0
• uvicorn==0.13.3
• Werkzeug==1.0.1

Project Description

The files and folders in this project and their usage is as follows.

• models: Contains pytorch model architecture used in this project. These models are loaded and updated using the saved checkpoints to run inference.

• sound_samples: Containes samples to run inference.

• static: Contains static files, css files, images to display and js files.

• templates: Contains templates for web app.

• uploads: Folder to store uploaded files.

• utils: Utility files for various supporting functions. Contains files for creating spectrograms, parameters for inference, data file to pre and post process audio files.

• app.py app file

• main.py File containing all of the functionality in the app. It imports certain other files to run properly.

• Inference.py Script to run inference o the audio clips.

Local Run

To run the app locally run the following

git clone [email protected]:speakupai/ml_deployment.git
cd ml_deployment

$ python app.py

The app should look like this