https://github.com/thomd/machine-learning-with-fastapi-on-heroku

Train a machine learning model for FastAPI and deploy on Heroku
https://github.com/thomd/machine-learning-with-fastapi-on-heroku

heroku machine-learning python

Last synced: 29 days ago
JSON representation

Train a machine learning model for FastAPI and deploy on Heroku

• Host: GitHub
• URL: https://github.com/thomd/machine-learning-with-fastapi-on-heroku
• Owner: thomd
• Created: 2021-11-20T14:39:16.000Z (over 4 years ago)
• Default Branch: main
• Last Pushed: 2021-11-20T23:46:49.000Z (over 4 years ago)
• Last Synced: 2025-02-05T21:42:52.119Z (over 1 year ago)
• Topics: heroku, machine-learning, python
• Language: Jupyter Notebook
• Homepage:
• Size: 10.1 MB
• Stars: 0
• Watchers: 2
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

          
# Train ML Model for FastAPI and Deploy on Heroku

* [FastAPI](https://fastapi.tiangolo.com) is a Web framework for developing RESTful APIs in Python.

* [Heroku](https://www.heroku.com) is a cloud platform as a service supporting several programming languages.

## Setup

Create Environment

    conda env create --file environment.yaml

    conda activate ml-fastapi-heroku

## 1. Data Analysis and Cleaning

Data is from the [Pima Indians Diabetes Database](https://data.world/data-society/pima-indians-diabetes-database).

Analyse and clean the data: [data-analysis.ipynb](./data-analysis.ipynb)

Compare original dataset with cleaned dataset with

    npx daff --www data/original/diabetes.csv data/diabetes.csv

## 2. Train Prediction Model

Traing of a model using a **random forest classifier**: [train-model.ipynb](./train-model.ipynb)

Test model with:

    import pickle

    model = pickle.load(open('model/model.pkl', 'rb'))

    p = model.predict_proba([[7,100,72,23,30.5,30.0,0.484,32]])

    print(p)

## 3. FastAPI Application

For local testing, start server with

    uvicorn app:app --reload

and open Swagger-UI:

    open http://localhost:8000/docs

## 4. Deploy on Heroku

    pip list --format=freeze > requirements.txt

    heroku login

    heroku create predict-diabetes-1

    git push heroku main

    heroku ps:scale web=1

    heroku logs --tail

    heroku open

## 5. Use API

    curl -s 'https://predict-diabetes-1.herokuapp.com/predict' \

         -H 'Content-Type: application/json' \

         -d '{"pregnancies":1,"glucose":89,"bp":66,"skinthickness":23,"insulin":94,"bmi":43,"dpf":0.167,"age":21}'