Pokemon Classifier based on fast.ai library deployed on Heroku.
This repo can be used as a starting point to deploy fast.ai models on Render.
The sample app described here is up at https://pokemonclassifierapp.herokuapp.com/. Test it out with pikachu,charmander or bulbasaur images!
You can check the notebook here.
-
The starter codes is taken from the starter code for deploying fast.ai models on Render.
-
We need to add a Procfile to the repository and put
web: python app/server.py serve
in it. -
We need to add a runtime.txt file to specify
python
version. -
In the
server.py
file we need to add :
import os
import requests
Port = int(os.environ.get('PORT', 50000))
and replace uvicorn.run(.....) by
uvicorn.run(app=app, host='0.0.0.0', port=Port, log_level="info")
Edit : After upgrading uvicorn version, the app crashed with the following error - Publish to Heroku is broken: "WARNING: You must pass the application as an import string to enable 'reload' or 'workers"
Solution is changing WEB_CONCURRENCY = 1 in app's setting on Heroku dashboard.