rliessum/docker-fastapi-ml

An example app that shows how to build a ML app with FastAPI and Docker.

The starter file is model_starter.py with a model from Hugging Face: https://huggingface.co/dandelin/vilt-b32-finetuned-vqa

The final app is in main.py and final model code is in model.py.

Below are the steps to use Docker:

Steps to Setup Docker and VS Code

STEP: Run docker init

docker init

STEP: Select Python image and use this command during docker init

CMD uvicorn main:app --reload --host 0.0.0.0 --port 8000

STEP: Remove -slim and add this to Dockerfile (Needed to install transformers)

# Install Rust compiler
RUN curl https://sh.rustup.rs -sSf | bash -s -- -y
ENV PATH="/root/.cargo/bin:${PATH}"

STEP: Remove the custom user from the Dockerfile to simplify, because HF needs write permissions to download and save the model.

STEP: Add volume to compose.yaml:

    volumes:
      - .:/app

STEP Run docker compose up --build

STEP: Attach VS Code to running container

Examples

• file: cats.jpeg

  text: How many cats are there?

• file: tiger.jpeg

  text: What's the animal doing?

• file: palace.jpeg

  text: What is on top of the building?