AaronWard/fastapi-model-deployment

Repo for FastAPI

---

This repo contains code for training and deploying a model for classifying whether a person has a salary above or below 50,000. This model is trained on census dats containing certain demographic data (More information here: https://archive.ics.uci.edu/ml/datasets/census+income)

Requirements

• Run pip install requirements.txt

Modelling

• Model consists of a Random Forest classifier which predicts whether someones salary is above $50,000. For this use case a binary classification approach was taken, whereby a sample showing a probability of 1.0 being positive and 0.0 being negative case.
• run ./starter/train_model.py to train the model.
• Write model results to model_card_template.md to persist details about the model

API Testing

• For local tests
  • start the API using uvicorn main:app --reload
  • query "http://127.0.0.1:8000" using the live_endpoint_query_script.py
  • You can update the values passed to the model to get differement model outputs.
• For live testing use the
  • query "https://award40-udacity.herokuapp.com" using the live_endpoint_query_script.py

Testing

• There are two unit test scripts: test_app.py and test_model.py
• test_app.py contains tests relating to the FastAPI.
• test_model.py contains tests ensuring that the code used to preprocess and train the model is working as intended
• Simply running pytest in the root folder via your terminal will allow to run the unit tests.