IA Example

Middle API using TensorFlow Serving.

Documentation

See web documents accessing to /doc or /redoc

Single entry-point /api/v1/resnet

QA

You could test the code executing pytest --cov=main or docker run -h --rm -i -t ia_example:latest -c pytest --cov=main, also the code quality with flake8

Build

Create docker image

• With Docker

docker build . --tag ia_example:latest

• With Docker-compose

docker-compose build

Docker Environment

• TENSORFLOW_PORT(By default 8500)

  Allows to change the TensorFlow port

• TENSORFLOW_API_REST_PORT(By default 8501)

  Allows to change the TensorFlow Rest API port

• TENSORFLOW_HOST(By default localhost)

  In order to allow internal or external TensorFlow services you could use that var to change the host used to process the prediction

• TENSORFLOW_ARGS(Empty by default)

  If TENSORFLOW_HOST="localhost" the args passed to tensorflow_model_server bin

• API_ARGS (Empty by default)

  Args used to uvicorn

Deploy

• With Docker

docker run -d -p 8000:8000 --name ia_example -e MODEL_NAME=resnet -t ia_example:latest

• With Docker-compose
  • NOTE : In docker compose will be deployed two containers, one with the API other with TensorFlow. Also include Prometheus monitoring system.

docker-compose up -d