Deploying ML models on Fly in a couple of steps

How To
1

Hi everyone! I put together a blog post on how to easily deploy an ML model to fly.io (much faster than to build a service and run flyctl deploy manually). In a nutshell, you just need to serialize your model (all Python requirements are found automatically, API interfaces are constructed) and then run a deployment command.

The blog post is about CV models, but it work with any ML model (or just Python function even - that expands to non-ML cases as well). You can deploy it as a FastAPI or Streamlit app (both are built automatically).

Hope you’ll find it useful! Happy to answer questions and help out in case something doesn’t work!

2 Likes
2

You can even save a trained machine learning model right from a Jupyter notebook or Colab and deploy it from there - and it just works

1 Like