How to create simple API service for Wav2Vec2 model using FastAPI + Docker

This documentation demonstrates how to achieve an API service for a deep learning model. Which is rely on Hugging Face's Transformer library to load the processor and model. Let's get started!

Prerequisite

We must install all of the following on the host machine to create API with docker and enable the GPU.

• Docker Select your OS and following the instruction.
• Docker Compose follow the instruction.
• Nvidia Drivers Select your GPU and OS specification in the drop-down menu. Then download and install it following by instructions. But in case Windows OS + WSL2 with Ubuntu installed. You must install the Nvidia driver on your Windows system only. Don't install it in WSL2 because WSL2 basically sees the Nvidia driver in Windows.
• NVIDIA Container Toolkit Follow this documentation (But you can skip that setting up docker because we already installed that.)