FastAPI TensorFlow Serving

GitHub project for Tensorflow-based Object Detection on the CIFAR-10 dataset, served with FastAPI.

Tech Stack

• Tensorflow 2
• FastAPI
• PDM
• Docker

Up and Running

Developer Setup

1. Create and activate a virtualenv with conda:

conda create -n tfserve python=3.11 -y
conda activate tfserve

2. Install the dependencies on pyproject.toml:

pdm sync --no-self

3. Start the app with gunicorn with:

uvicorn --app-dir app/ main:app --host 0.0.0.0 --port 8000 --reload

or simply execute:

make run

4. Access the Swagger UI at:

open http://localhost:8000

Containerization and Testing

1. Build the Docker Image with:

make docker-image

2. Spin up the container with:

make docker-run

3. Access the Swagger UI at:

open http://localhost:8000

TODO

• PEP-517: Packaging and dependency management with PDM
• Code format/lint with Ruff
• Run on Docker
• Serve it with Streamlit
• GitHub CI