Deploy FastAPI+PyTorch API with Docker

docker-compose.yml

# Example for Multi-GPU deploying

services:
  device0:
    image: image/name
    deploy:
      resources:
        reservations:
          devices:
          - driver: nvidia
            device_ids: ['0']
            capabilities: [gpu]
    ports:
      - "5000:3000"

  device1:
    image: image/name
    deploy:
      resources:
        reservations:
          devices:
          - driver: nvidia
            device_ids: ['1']
            capabilities: [gpu]
    ports:
      - "5001:3000"

  device2:
    image: image/name
    deploy:
      resources:
        reservations:
          devices:
          - driver: nvidia
            device_ids: ['2']
            capabilities: [gpu]
    ports:
      - "5002:3000"

  device3:
    image: image/name
    deploy:
      resources:
        reservations:
          devices:
          - driver: nvidia
            device_ids: ['3']
            capabilities: [gpu]
    ports:
      - "5003:3000"

Dockerfile

## First, clone your api repository to current folder. and copy Dockerfile
FROM pytorch/pytorch:latest

## Copying current folder to image
COPY ./ ./

## install git-lfs for downloading Huggingface model and datasets. 
RUN apt update -y && apt install git git-lfs -y
RUN git lfs install

## Download model 
RUN git clone https://huggingface.co/{Model Name}

## Download Dataset
RUN git clone https://huggingface.co/{Dataset Name}

## Install dependencies
RUN pip install --upgrade pip
RUN pip install -r requirements.txt

## Exposing port 
EXPOSE 3000

## Running command 
# Example for FastAPI
CMD ["uvicorn", "app:app", "--port=3000", "--host=0.0.0.0"]